AgMoDB
ModelsAgentsOrchestratorsEvalsCompositesVisualizeIndustry
AgMoDB by @mistakeknot
All Models

GPT-6 Sol

Closed-weight OpenAI mid-tier GPT-6 model for demanding everyday and coding work, positioned between GPT-6 Astra and GPT-6 Luna, with effort-tiered reasoning.

September 22, 2026Last synced Nov 22, 2026, 8:17 AM1050K context
Effort
  • No reasoning
  • Default
  • Low
  • Medium
  • High
  • Xhigh

Blended Price

$4.00/M

Input Price

$2.00/M

Output Price

$10.00/M

Speed

—

TTFT

—

Benchmark Scores

Intelligence Index
44.2
Coding Index
—
Math Index
—
MMLU Pro
— / 100
GPQA Diamond
— / 100
HLE
47.9 / 100
LiveCodeBench
— / 100
SciCode
57.6 / 100
MATH-500
— / 100
AIME
— / 30
AIME 2025
— / 30
IFBench
— / 100
LCR
83.7 / 100
Terminal-Bench Hard
— / 100
τ²-Bench
— / 100

Performance Radar

GPT-6 Sol
Field Average

Hidden, no data: Output Speed

External Benchmarks

AA-Omniscience Accuracyaa_omniscience
54.5 / 100
AA-Omniscience Hallucination Rateaa_omniscience
60.1 / 100
AA-Omniscience Indexaa_omniscience
27.1 / 100
ARC-AGI-1arc_agi
29.3 / 100
ARC-AGI-1 Cost per Taskarc_agi
0.0
ARC-AGI-2arc_agi
1.7 / 100
ARC-AGI-2 Cost per Taskarc_agi
0.1
BullshitBenchbullshitbench
57.0 / 100
Arena ELO: Codingchatbot_arena
1674.0
Chatbot Arena ELOchatbot_arena
1398.0
arena_elo_searchchatbot_arena
1174.0
arena_elo_visionchatbot_arena
1232.0
Ale Benchepoch_ai
2462.0
Apex Agentsepoch_ai
0.5
Arc Agi 2epoch_ai
0.0
Blueprint Bench 2epoch_ai
0.4
Chess Puzzlesepoch_ai
0.6
Critptepoch_ai
0.0
Dtbenchepoch_ai
1.0
Ebr Benchepoch_ai
0.5
Frontiercodeepoch_ai
0.5
Frontiermath Erdosepoch_ai
0.0
Frontiermath Tier 4 V2epoch_ai
0.9
Frontiermath Tiers 1 3 V2epoch_ai
0.9
Furniture Assemblyepoch_ai
0.6
Gdp Pdfepoch_ai
0.3
Lmcaepoch_ai
59.1
Mirrorcodeepoch_ai
0.5
Mystery Game Puzzlesepoch_ai
0.6
OTIS Mock AIME 2024–2025epoch_ai
1.0
Proofbenchepoch_ai
0.8
Scicodeepoch_ai
0.5
SimpleQA Verified (Epoch)epoch_ai
0.6
Spireclimbepoch_ai
0.1
Vending Bench 2epoch_ai
14427.8
WebDev Arenaepoch_ai
1689.3
LiveBench Codinglivebench
93.1 / 100
LiveBench Data Analysislivebench
81.2 / 100
LiveBench Instruction Followinglivebench
68.6 / 100
LiveBench Languagelivebench
85.3 / 100
LiveBench Mathlivebench
96.4 / 100
LiveBench Reasoninglivebench
88.7 / 100
SimpleBenchsimplebench
73.1 / 100
Vectara Factual Consistencyvectara_hallucination
93.5 / 100
Vectara Hallucination Ratevectara_hallucination
6.5 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-medium-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'medium' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"ec09fbed9f211f59ec54cca2227c3309\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"5814b0938ba4a54ba31da46fe17d4380\"",
  "upstreamModelId": "openai-gpt-6-sol-medium-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5889.174868,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-20

4.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-high-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"ec09fbed9f211f59ec54cca2227c3309\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"5814b0938ba4a54ba31da46fe17d4380\"",
  "upstreamModelId": "openai-gpt-6-sol-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5908.428836,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-20

5.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the standard harness.",
  "modelsEtag": "W/\"ec09fbed9f211f59ec54cca2227c3309\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"5814b0938ba4a54ba31da46fe17d4380\"",
  "upstreamModelId": "openai-gpt-6-sol-max",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5554.493172,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-20

4.6 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"ec09fbed9f211f59ec54cca2227c3309\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"5814b0938ba4a54ba31da46fe17d4380\"",
  "upstreamModelId": "openai-gpt-6-sol-max-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 8721.949674,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-20

23.0 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (XHigh) · effort xhigh · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-xhigh-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'xhigh' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"ec09fbed9f211f59ec54cca2227c3309\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"5814b0938ba4a54ba31da46fe17d4380\"",
  "upstreamModelId": "openai-gpt-6-sol-xhigh-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 6906.642396,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-20

9.5 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the standard harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts.",
  "modelsEtag": "W/\"ec09fbed9f211f59ec54cca2227c3309\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"5814b0938ba4a54ba31da46fe17d4380\"",
  "upstreamModelId": "openai-gpt-6-sol-none",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4655.549962,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-20

0.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"ec09fbed9f211f59ec54cca2227c3309\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"5814b0938ba4a54ba31da46fe17d4380\"",
  "upstreamModelId": "openai-gpt-6-sol-none-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4233.770312,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-20

1.2 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-low-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'low' reasoning effort using the provider adapter harness.",
  "modelsEtag": "W/\"ec09fbed9f211f59ec54cca2227c3309\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"5814b0938ba4a54ba31da46fe17d4380\"",
  "upstreamModelId": "openai-gpt-6-sol-low-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4745.286212,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-20

1.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"d8dfdc9958a2e1502b54603b03fbc557\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"aaa10807f3dcaae3ce6c15be7539e733\"",
  "upstreamModelId": "openai-gpt-6-sol-max-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 8721.949674,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-20

23.0 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (XHigh) · effort xhigh · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-xhigh-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'xhigh' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"d8dfdc9958a2e1502b54603b03fbc557\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"aaa10807f3dcaae3ce6c15be7539e733\"",
  "upstreamModelId": "openai-gpt-6-sol-xhigh-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 6906.642396,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-20

9.5 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-high-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"d8dfdc9958a2e1502b54603b03fbc557\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"aaa10807f3dcaae3ce6c15be7539e733\"",
  "upstreamModelId": "openai-gpt-6-sol-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5908.428836,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-20

5.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-medium-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'medium' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"d8dfdc9958a2e1502b54603b03fbc557\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"aaa10807f3dcaae3ce6c15be7539e733\"",
  "upstreamModelId": "openai-gpt-6-sol-medium-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5889.174868,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-20

4.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-low-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'low' reasoning effort using the provider adapter harness.",
  "modelsEtag": "W/\"d8dfdc9958a2e1502b54603b03fbc557\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"aaa10807f3dcaae3ce6c15be7539e733\"",
  "upstreamModelId": "openai-gpt-6-sol-low-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4745.286212,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-20

1.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the standard harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts.",
  "modelsEtag": "W/\"d8dfdc9958a2e1502b54603b03fbc557\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"aaa10807f3dcaae3ce6c15be7539e733\"",
  "upstreamModelId": "openai-gpt-6-sol-none",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4655.549962,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-20

0.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the standard harness.",
  "modelsEtag": "W/\"d8dfdc9958a2e1502b54603b03fbc557\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"aaa10807f3dcaae3ce6c15be7539e733\"",
  "upstreamModelId": "openai-gpt-6-sol-max",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5554.493172,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-20

4.6 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"d8dfdc9958a2e1502b54603b03fbc557\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"aaa10807f3dcaae3ce6c15be7539e733\"",
  "upstreamModelId": "openai-gpt-6-sol-none-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4233.770312,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-20

1.2 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-low-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'low' reasoning effort using the provider adapter harness.",
  "modelsEtag": "W/\"f4d8307b32b2fefe010860bedd0c5dd8\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"ef15ebd7644a875688eda094d716da9a\"",
  "upstreamModelId": "openai-gpt-6-sol-low-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4745.286212,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-18

1.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"f4d8307b32b2fefe010860bedd0c5dd8\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"ef15ebd7644a875688eda094d716da9a\"",
  "upstreamModelId": "openai-gpt-6-sol-none-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4233.770312,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-18

1.2 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the standard harness.",
  "modelsEtag": "W/\"f4d8307b32b2fefe010860bedd0c5dd8\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"ef15ebd7644a875688eda094d716da9a\"",
  "upstreamModelId": "openai-gpt-6-sol-max",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5554.493172,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-18

4.6 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"f4d8307b32b2fefe010860bedd0c5dd8\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"ef15ebd7644a875688eda094d716da9a\"",
  "upstreamModelId": "openai-gpt-6-sol-max-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 8721.949674,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-18

23.0 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-high-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"f4d8307b32b2fefe010860bedd0c5dd8\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"ef15ebd7644a875688eda094d716da9a\"",
  "upstreamModelId": "openai-gpt-6-sol-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5908.428836,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-18

5.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-medium-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'medium' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"f4d8307b32b2fefe010860bedd0c5dd8\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"ef15ebd7644a875688eda094d716da9a\"",
  "upstreamModelId": "openai-gpt-6-sol-medium-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5889.174868,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-18

4.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the standard harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts.",
  "modelsEtag": "W/\"f4d8307b32b2fefe010860bedd0c5dd8\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"ef15ebd7644a875688eda094d716da9a\"",
  "upstreamModelId": "openai-gpt-6-sol-none",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4655.549962,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-18

0.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (XHigh) · effort xhigh · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-xhigh-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'xhigh' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"f4d8307b32b2fefe010860bedd0c5dd8\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"ef15ebd7644a875688eda094d716da9a\"",
  "upstreamModelId": "openai-gpt-6-sol-xhigh-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 6906.642396,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-18

9.5 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the standard harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts.",
  "modelsEtag": "W/\"f4d8307b32b2fefe010860bedd0c5dd8\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"ef15ebd7644a875688eda094d716da9a\"",
  "upstreamModelId": "openai-gpt-6-sol-none",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4655.549962,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-17

0.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-low-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'low' reasoning effort using the provider adapter harness.",
  "modelsEtag": "W/\"f4d8307b32b2fefe010860bedd0c5dd8\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"ef15ebd7644a875688eda094d716da9a\"",
  "upstreamModelId": "openai-gpt-6-sol-low-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4745.286212,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-17

1.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the standard harness.",
  "modelsEtag": "W/\"f4d8307b32b2fefe010860bedd0c5dd8\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"ef15ebd7644a875688eda094d716da9a\"",
  "upstreamModelId": "openai-gpt-6-sol-max",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5554.493172,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-17

4.6 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"f4d8307b32b2fefe010860bedd0c5dd8\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"ef15ebd7644a875688eda094d716da9a\"",
  "upstreamModelId": "openai-gpt-6-sol-none-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4233.770312,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-17

1.2 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-medium-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'medium' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"f4d8307b32b2fefe010860bedd0c5dd8\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"ef15ebd7644a875688eda094d716da9a\"",
  "upstreamModelId": "openai-gpt-6-sol-medium-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5889.174868,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-17

4.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-high-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"f4d8307b32b2fefe010860bedd0c5dd8\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"ef15ebd7644a875688eda094d716da9a\"",
  "upstreamModelId": "openai-gpt-6-sol-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5908.428836,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-17

5.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (XHigh) · effort xhigh · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-xhigh-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'xhigh' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"f4d8307b32b2fefe010860bedd0c5dd8\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"ef15ebd7644a875688eda094d716da9a\"",
  "upstreamModelId": "openai-gpt-6-sol-xhigh-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 6906.642396,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-17

9.5 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"f4d8307b32b2fefe010860bedd0c5dd8\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"ef15ebd7644a875688eda094d716da9a\"",
  "upstreamModelId": "openai-gpt-6-sol-max-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 8721.949674,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-17

23.0 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"69aae253e532553a920bf4deb50ca495\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"5206202f381a21893de82dcc2db93ae9\"",
  "upstreamModelId": "openai-gpt-6-sol-max-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 8721.949674,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-16

23.0 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (XHigh) · effort xhigh · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-xhigh-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'xhigh' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"69aae253e532553a920bf4deb50ca495\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"5206202f381a21893de82dcc2db93ae9\"",
  "upstreamModelId": "openai-gpt-6-sol-xhigh-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 6906.642396,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-16

9.5 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the standard harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts.",
  "modelsEtag": "W/\"69aae253e532553a920bf4deb50ca495\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"5206202f381a21893de82dcc2db93ae9\"",
  "upstreamModelId": "openai-gpt-6-sol-none",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4655.549962,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-16

0.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-high-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"69aae253e532553a920bf4deb50ca495\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"5206202f381a21893de82dcc2db93ae9\"",
  "upstreamModelId": "openai-gpt-6-sol-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5908.428836,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-16

5.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the standard harness.",
  "modelsEtag": "W/\"69aae253e532553a920bf4deb50ca495\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"5206202f381a21893de82dcc2db93ae9\"",
  "upstreamModelId": "openai-gpt-6-sol-max",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5554.493172,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-16

4.6 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-medium-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'medium' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"69aae253e532553a920bf4deb50ca495\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"5206202f381a21893de82dcc2db93ae9\"",
  "upstreamModelId": "openai-gpt-6-sol-medium-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5889.174868,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-16

4.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"69aae253e532553a920bf4deb50ca495\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"5206202f381a21893de82dcc2db93ae9\"",
  "upstreamModelId": "openai-gpt-6-sol-none-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4233.770312,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-16

1.2 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-low-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'low' reasoning effort using the provider adapter harness.",
  "modelsEtag": "W/\"69aae253e532553a920bf4deb50ca495\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"5206202f381a21893de82dcc2db93ae9\"",
  "upstreamModelId": "openai-gpt-6-sol-low-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4745.286212,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-16

1.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"513aaaa353972ca0baa9c95cb95e04dd\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"169653a17d103c8dbca682167ffccaa6\"",
  "upstreamModelId": "openai-gpt-6-sol-max-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 8721.949674,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-10

23.0 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the standard harness.",
  "modelsEtag": "W/\"513aaaa353972ca0baa9c95cb95e04dd\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"169653a17d103c8dbca682167ffccaa6\"",
  "upstreamModelId": "openai-gpt-6-sol-max",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5554.493172,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-10

4.6 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-low-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'low' reasoning effort using the provider adapter harness.",
  "modelsEtag": "W/\"513aaaa353972ca0baa9c95cb95e04dd\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"169653a17d103c8dbca682167ffccaa6\"",
  "upstreamModelId": "openai-gpt-6-sol-low-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4745.286212,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-10

1.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"513aaaa353972ca0baa9c95cb95e04dd\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"169653a17d103c8dbca682167ffccaa6\"",
  "upstreamModelId": "openai-gpt-6-sol-none-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4233.770312,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-10

1.2 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-medium-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'medium' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"513aaaa353972ca0baa9c95cb95e04dd\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"169653a17d103c8dbca682167ffccaa6\"",
  "upstreamModelId": "openai-gpt-6-sol-medium-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5889.174868,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-10

4.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-high-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"513aaaa353972ca0baa9c95cb95e04dd\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"169653a17d103c8dbca682167ffccaa6\"",
  "upstreamModelId": "openai-gpt-6-sol-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5908.428836,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-10

5.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (XHigh) · effort xhigh · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-xhigh-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'xhigh' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"513aaaa353972ca0baa9c95cb95e04dd\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"169653a17d103c8dbca682167ffccaa6\"",
  "upstreamModelId": "openai-gpt-6-sol-xhigh-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 6906.642396,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-10

9.5 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the standard harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts.",
  "modelsEtag": "W/\"513aaaa353972ca0baa9c95cb95e04dd\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"169653a17d103c8dbca682167ffccaa6\"",
  "upstreamModelId": "openai-gpt-6-sol-none",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4655.549962,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-10

0.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-low-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'low' reasoning effort using the provider adapter harness.",
  "modelsEtag": "W/\"3d7dcca4fb88130b9e7ec186fd9ae23f\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"1e46e6ab8c83fb1a7b19f10be398adf2\"",
  "upstreamModelId": "openai-gpt-6-sol-low-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4745.286212,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-09

1.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"3d7dcca4fb88130b9e7ec186fd9ae23f\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"1e46e6ab8c83fb1a7b19f10be398adf2\"",
  "upstreamModelId": "openai-gpt-6-sol-none-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4233.770312,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-09

1.2 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (XHigh) · effort xhigh · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-xhigh-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'xhigh' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"3d7dcca4fb88130b9e7ec186fd9ae23f\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"1e46e6ab8c83fb1a7b19f10be398adf2\"",
  "upstreamModelId": "openai-gpt-6-sol-xhigh-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 6906.642396,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-09

9.5 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the standard harness.",
  "modelsEtag": "W/\"3d7dcca4fb88130b9e7ec186fd9ae23f\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"1e46e6ab8c83fb1a7b19f10be398adf2\"",
  "upstreamModelId": "openai-gpt-6-sol-max",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5554.493172,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-09

4.6 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-medium-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'medium' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"3d7dcca4fb88130b9e7ec186fd9ae23f\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"1e46e6ab8c83fb1a7b19f10be398adf2\"",
  "upstreamModelId": "openai-gpt-6-sol-medium-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5889.174868,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-09

4.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-high-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"3d7dcca4fb88130b9e7ec186fd9ae23f\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"1e46e6ab8c83fb1a7b19f10be398adf2\"",
  "upstreamModelId": "openai-gpt-6-sol-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5908.428836,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-09

5.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"3d7dcca4fb88130b9e7ec186fd9ae23f\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"1e46e6ab8c83fb1a7b19f10be398adf2\"",
  "upstreamModelId": "openai-gpt-6-sol-max-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 8721.949674,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-09

23.0 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the standard harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts.",
  "modelsEtag": "W/\"3d7dcca4fb88130b9e7ec186fd9ae23f\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"1e46e6ab8c83fb1a7b19f10be398adf2\"",
  "upstreamModelId": "openai-gpt-6-sol-none",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4655.549962,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-09

0.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-low-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'low' reasoning effort using the provider adapter harness.",
  "modelsEtag": "W/\"8ea4d2292a0a8c759aa7e80a2041122d\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"94fd3d68aacca744ff3c39f63a187685\"",
  "upstreamModelId": "openai-gpt-6-sol-low-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4745.286212,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-06

1.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"8ea4d2292a0a8c759aa7e80a2041122d\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"94fd3d68aacca744ff3c39f63a187685\"",
  "upstreamModelId": "openai-gpt-6-sol-none-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4233.770312,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-06

1.2 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-high-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"8ea4d2292a0a8c759aa7e80a2041122d\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"94fd3d68aacca744ff3c39f63a187685\"",
  "upstreamModelId": "openai-gpt-6-sol-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5908.428836,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-06

5.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (XHigh) · effort xhigh · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-xhigh-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'xhigh' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"8ea4d2292a0a8c759aa7e80a2041122d\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"94fd3d68aacca744ff3c39f63a187685\"",
  "upstreamModelId": "openai-gpt-6-sol-xhigh-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 6906.642396,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-06

9.5 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"8ea4d2292a0a8c759aa7e80a2041122d\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"94fd3d68aacca744ff3c39f63a187685\"",
  "upstreamModelId": "openai-gpt-6-sol-max-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 8721.949674,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-06

23.0 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the standard harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts.",
  "modelsEtag": "W/\"8ea4d2292a0a8c759aa7e80a2041122d\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"94fd3d68aacca744ff3c39f63a187685\"",
  "upstreamModelId": "openai-gpt-6-sol-none",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4655.549962,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-06

0.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-medium-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'medium' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"8ea4d2292a0a8c759aa7e80a2041122d\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"94fd3d68aacca744ff3c39f63a187685\"",
  "upstreamModelId": "openai-gpt-6-sol-medium-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5889.174868,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-06

4.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the standard harness.",
  "modelsEtag": "W/\"8ea4d2292a0a8c759aa7e80a2041122d\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"94fd3d68aacca744ff3c39f63a187685\"",
  "upstreamModelId": "openai-gpt-6-sol-max",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5554.493172,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-06

4.6 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (XHigh) · effort xhigh · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-xhigh-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'xhigh' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"c4baed683b2f33c40537f2338682475f\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"48d1fd5bf96ae9583213f6699f12f2cc\"",
  "upstreamModelId": "openai-gpt-6-sol-xhigh-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 6906.642396,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-04

9.5 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-low-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'low' reasoning effort using the provider adapter harness.",
  "modelsEtag": "W/\"c4baed683b2f33c40537f2338682475f\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"48d1fd5bf96ae9583213f6699f12f2cc\"",
  "upstreamModelId": "openai-gpt-6-sol-low-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4745.286212,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-04

1.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-medium-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'medium' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"c4baed683b2f33c40537f2338682475f\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"48d1fd5bf96ae9583213f6699f12f2cc\"",
  "upstreamModelId": "openai-gpt-6-sol-medium-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5889.174868,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-04

4.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-high-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"c4baed683b2f33c40537f2338682475f\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"48d1fd5bf96ae9583213f6699f12f2cc\"",
  "upstreamModelId": "openai-gpt-6-sol-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5908.428836,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-04

5.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the standard harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts.",
  "modelsEtag": "W/\"c4baed683b2f33c40537f2338682475f\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"48d1fd5bf96ae9583213f6699f12f2cc\"",
  "upstreamModelId": "openai-gpt-6-sol-none",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4655.549962,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-04

0.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the standard harness.",
  "modelsEtag": "W/\"c4baed683b2f33c40537f2338682475f\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"48d1fd5bf96ae9583213f6699f12f2cc\"",
  "upstreamModelId": "openai-gpt-6-sol-max",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5554.493172,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-04

4.6 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"c4baed683b2f33c40537f2338682475f\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"48d1fd5bf96ae9583213f6699f12f2cc\"",
  "upstreamModelId": "openai-gpt-6-sol-max-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 8721.949674,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-04

23.0 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"c4baed683b2f33c40537f2338682475f\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"48d1fd5bf96ae9583213f6699f12f2cc\"",
  "upstreamModelId": "openai-gpt-6-sol-none-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4233.770312,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-04

1.2 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-high-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"c4baed683b2f33c40537f2338682475f\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"48d1fd5bf96ae9583213f6699f12f2cc\"",
  "upstreamModelId": "openai-gpt-6-sol-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5908.428836,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-02

5.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the standard harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts.",
  "modelsEtag": "W/\"c4baed683b2f33c40537f2338682475f\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"48d1fd5bf96ae9583213f6699f12f2cc\"",
  "upstreamModelId": "openai-gpt-6-sol-none",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4655.549962,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-02

0.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"c4baed683b2f33c40537f2338682475f\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"48d1fd5bf96ae9583213f6699f12f2cc\"",
  "upstreamModelId": "openai-gpt-6-sol-max-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 8721.949674,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-02

23.0 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-low-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'low' reasoning effort using the provider adapter harness.",
  "modelsEtag": "W/\"c4baed683b2f33c40537f2338682475f\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"48d1fd5bf96ae9583213f6699f12f2cc\"",
  "upstreamModelId": "openai-gpt-6-sol-low-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4745.286212,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-02

1.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (XHigh) · effort xhigh · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-xhigh-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'xhigh' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"c4baed683b2f33c40537f2338682475f\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"48d1fd5bf96ae9583213f6699f12f2cc\"",
  "upstreamModelId": "openai-gpt-6-sol-xhigh-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 6906.642396,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-02

9.5 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-medium-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'medium' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"c4baed683b2f33c40537f2338682475f\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"48d1fd5bf96ae9583213f6699f12f2cc\"",
  "upstreamModelId": "openai-gpt-6-sol-medium-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5889.174868,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-02

4.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"c4baed683b2f33c40537f2338682475f\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"48d1fd5bf96ae9583213f6699f12f2cc\"",
  "upstreamModelId": "openai-gpt-6-sol-none-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4233.770312,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-02

1.2 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the standard harness.",
  "modelsEtag": "W/\"c4baed683b2f33c40537f2338682475f\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"48d1fd5bf96ae9583213f6699f12f2cc\"",
  "upstreamModelId": "openai-gpt-6-sol-max",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5554.493172,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-02

4.6 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-low-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'low' reasoning effort using the provider adapter harness.",
  "modelsEtag": "W/\"8fc20a5217998d26c40ad2306d648194\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"095b2fa20aac96121092dab07e18cc8e\"",
  "upstreamModelId": "openai-gpt-6-sol-low-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4745.286212,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-28

1.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (XHigh) · effort xhigh · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-xhigh-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'xhigh' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"8fc20a5217998d26c40ad2306d648194\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"095b2fa20aac96121092dab07e18cc8e\"",
  "upstreamModelId": "openai-gpt-6-sol-xhigh-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 6906.642396,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-28

9.5 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-medium-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'medium' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"8fc20a5217998d26c40ad2306d648194\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"095b2fa20aac96121092dab07e18cc8e\"",
  "upstreamModelId": "openai-gpt-6-sol-medium-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5889.174868,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-28

4.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"8fc20a5217998d26c40ad2306d648194\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"095b2fa20aac96121092dab07e18cc8e\"",
  "upstreamModelId": "openai-gpt-6-sol-max-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 8721.949674,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-28

23.0 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"8fc20a5217998d26c40ad2306d648194\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"095b2fa20aac96121092dab07e18cc8e\"",
  "upstreamModelId": "openai-gpt-6-sol-none-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4233.770312,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-28

1.2 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the standard harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts.",
  "modelsEtag": "W/\"8fc20a5217998d26c40ad2306d648194\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"095b2fa20aac96121092dab07e18cc8e\"",
  "upstreamModelId": "openai-gpt-6-sol-none",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4655.549962,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-28

0.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-high-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"8fc20a5217998d26c40ad2306d648194\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"095b2fa20aac96121092dab07e18cc8e\"",
  "upstreamModelId": "openai-gpt-6-sol-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5908.428836,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-28

5.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the standard harness.",
  "modelsEtag": "W/\"8fc20a5217998d26c40ad2306d648194\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"095b2fa20aac96121092dab07e18cc8e\"",
  "upstreamModelId": "openai-gpt-6-sol-max",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5554.493172,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-28

4.6 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-low-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'low' reasoning effort using the provider adapter harness.",
  "modelsEtag": "W/\"400b32b66e8c1b37fc50ff586b033048\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"9314d4bea0d9d001c69a11614dc8d93c\"",
  "upstreamModelId": "openai-gpt-6-sol-low-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4745.286212,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-21

1.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"400b32b66e8c1b37fc50ff586b033048\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"9314d4bea0d9d001c69a11614dc8d93c\"",
  "upstreamModelId": "openai-gpt-6-sol-max-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 8721.949674,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-21

23.0 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-high-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"400b32b66e8c1b37fc50ff586b033048\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"9314d4bea0d9d001c69a11614dc8d93c\"",
  "upstreamModelId": "openai-gpt-6-sol-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5908.428836,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-21

5.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-medium-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'medium' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"400b32b66e8c1b37fc50ff586b033048\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"9314d4bea0d9d001c69a11614dc8d93c\"",
  "upstreamModelId": "openai-gpt-6-sol-medium-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5889.174868,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-21

4.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the standard harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts.",
  "modelsEtag": "W/\"400b32b66e8c1b37fc50ff586b033048\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"9314d4bea0d9d001c69a11614dc8d93c\"",
  "upstreamModelId": "openai-gpt-6-sol-none",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4655.549962,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-21

0.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the standard harness.",
  "modelsEtag": "W/\"400b32b66e8c1b37fc50ff586b033048\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"9314d4bea0d9d001c69a11614dc8d93c\"",
  "upstreamModelId": "openai-gpt-6-sol-max",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5554.493172,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-21

4.6 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (XHigh) · effort xhigh · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-xhigh-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'xhigh' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"400b32b66e8c1b37fc50ff586b033048\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"9314d4bea0d9d001c69a11614dc8d93c\"",
  "upstreamModelId": "openai-gpt-6-sol-xhigh-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 6906.642396,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-21

9.5 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"400b32b66e8c1b37fc50ff586b033048\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"9314d4bea0d9d001c69a11614dc8d93c\"",
  "upstreamModelId": "openai-gpt-6-sol-none-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4233.770312,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-21

1.2 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-high-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"2ab08f641c16bcf3c7a59c38b35fa2e3\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"b4ed41aee6b00420557f098bd05f4165\"",
  "upstreamModelId": "openai-gpt-6-sol-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5908.428836,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-19

5.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"2ab08f641c16bcf3c7a59c38b35fa2e3\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"b4ed41aee6b00420557f098bd05f4165\"",
  "upstreamModelId": "openai-gpt-6-sol-none-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4233.770312,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-19

1.2 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the standard harness.",
  "modelsEtag": "W/\"2ab08f641c16bcf3c7a59c38b35fa2e3\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"b4ed41aee6b00420557f098bd05f4165\"",
  "upstreamModelId": "openai-gpt-6-sol-max",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5554.493172,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-19

4.6 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-low-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'low' reasoning effort using the provider adapter harness.",
  "modelsEtag": "W/\"2ab08f641c16bcf3c7a59c38b35fa2e3\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"b4ed41aee6b00420557f098bd05f4165\"",
  "upstreamModelId": "openai-gpt-6-sol-low-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4745.286212,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-19

1.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the standard harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts.",
  "modelsEtag": "W/\"2ab08f641c16bcf3c7a59c38b35fa2e3\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"b4ed41aee6b00420557f098bd05f4165\"",
  "upstreamModelId": "openai-gpt-6-sol-none",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4655.549962,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-19

0.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"2ab08f641c16bcf3c7a59c38b35fa2e3\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"b4ed41aee6b00420557f098bd05f4165\"",
  "upstreamModelId": "openai-gpt-6-sol-max-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 8721.949674,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-19

23.0 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (XHigh) · effort xhigh · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-xhigh-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'xhigh' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"2ab08f641c16bcf3c7a59c38b35fa2e3\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"b4ed41aee6b00420557f098bd05f4165\"",
  "upstreamModelId": "openai-gpt-6-sol-xhigh-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 6906.642396,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-19

9.5 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-medium-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'medium' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"2ab08f641c16bcf3c7a59c38b35fa2e3\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"b4ed41aee6b00420557f098bd05f4165\"",
  "upstreamModelId": "openai-gpt-6-sol-medium-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5889.174868,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-19

4.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (XHigh) · effort xhigh · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-xhigh-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'xhigh' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"8cf5cb76df0b13c1684b346a2012df83\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"c6e9a78d12dc070d7dab18a1f667a410\"",
  "upstreamModelId": "openai-gpt-6-sol-xhigh-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 6906.642396,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-15

9.5 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-low-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'low' reasoning effort using the provider adapter harness.",
  "modelsEtag": "W/\"8cf5cb76df0b13c1684b346a2012df83\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"c6e9a78d12dc070d7dab18a1f667a410\"",
  "upstreamModelId": "openai-gpt-6-sol-low-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4745.286212,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-15

1.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the standard harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts.",
  "modelsEtag": "W/\"8cf5cb76df0b13c1684b346a2012df83\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"c6e9a78d12dc070d7dab18a1f667a410\"",
  "upstreamModelId": "openai-gpt-6-sol-none",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4655.549962,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-15

0.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"8cf5cb76df0b13c1684b346a2012df83\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"c6e9a78d12dc070d7dab18a1f667a410\"",
  "upstreamModelId": "openai-gpt-6-sol-max-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 8721.949674,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-15

23.0 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-high-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"8cf5cb76df0b13c1684b346a2012df83\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"c6e9a78d12dc070d7dab18a1f667a410\"",
  "upstreamModelId": "openai-gpt-6-sol-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5908.428836,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-15

5.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-medium-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'medium' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"8cf5cb76df0b13c1684b346a2012df83\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"c6e9a78d12dc070d7dab18a1f667a410\"",
  "upstreamModelId": "openai-gpt-6-sol-medium-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5889.174868,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-15

4.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"8cf5cb76df0b13c1684b346a2012df83\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"c6e9a78d12dc070d7dab18a1f667a410\"",
  "upstreamModelId": "openai-gpt-6-sol-none-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4233.770312,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-15

1.2 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the standard harness.",
  "modelsEtag": "W/\"8cf5cb76df0b13c1684b346a2012df83\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"c6e9a78d12dc070d7dab18a1f667a410\"",
  "upstreamModelId": "openai-gpt-6-sol-max",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5554.493172,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-15

4.6 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (XHigh) · effort xhigh · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-xhigh-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'xhigh' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"57cc4a7689e563a0ffb7d275eb3a889b\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"19d71f8c9a2056c0be4bb4fd4a75a325\"",
  "upstreamModelId": "openai-gpt-6-sol-xhigh-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 6906.642396,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-14

9.5 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the standard harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts.",
  "modelsEtag": "W/\"57cc4a7689e563a0ffb7d275eb3a889b\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"19d71f8c9a2056c0be4bb4fd4a75a325\"",
  "upstreamModelId": "openai-gpt-6-sol-none",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4655.549962,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-14

0.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the standard harness.",
  "modelsEtag": "W/\"57cc4a7689e563a0ffb7d275eb3a889b\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"19d71f8c9a2056c0be4bb4fd4a75a325\"",
  "upstreamModelId": "openai-gpt-6-sol-max",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5554.493172,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-14

4.6 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"57cc4a7689e563a0ffb7d275eb3a889b\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"19d71f8c9a2056c0be4bb4fd4a75a325\"",
  "upstreamModelId": "openai-gpt-6-sol-max-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 8721.949674,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-14

23.0 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"57cc4a7689e563a0ffb7d275eb3a889b\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"19d71f8c9a2056c0be4bb4fd4a75a325\"",
  "upstreamModelId": "openai-gpt-6-sol-none-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4233.770312,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-14

1.2 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-low-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'low' reasoning effort using the provider adapter harness.",
  "modelsEtag": "W/\"57cc4a7689e563a0ffb7d275eb3a889b\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"19d71f8c9a2056c0be4bb4fd4a75a325\"",
  "upstreamModelId": "openai-gpt-6-sol-low-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4745.286212,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-14

1.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-medium-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'medium' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"57cc4a7689e563a0ffb7d275eb3a889b\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"19d71f8c9a2056c0be4bb4fd4a75a325\"",
  "upstreamModelId": "openai-gpt-6-sol-medium-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5889.174868,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-14

4.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-high-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"57cc4a7689e563a0ffb7d275eb3a889b\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"19d71f8c9a2056c0be4bb4fd4a75a325\"",
  "upstreamModelId": "openai-gpt-6-sol-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5908.428836,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-14

5.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-medium-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'medium' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"57cc4a7689e563a0ffb7d275eb3a889b\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"19d71f8c9a2056c0be4bb4fd4a75a325\"",
  "upstreamModelId": "openai-gpt-6-sol-medium-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5889.174868,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-14

4.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-high-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"57cc4a7689e563a0ffb7d275eb3a889b\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"19d71f8c9a2056c0be4bb4fd4a75a325\"",
  "upstreamModelId": "openai-gpt-6-sol-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5908.428836,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-14

5.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (XHigh) · effort xhigh · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-xhigh-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'xhigh' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"57cc4a7689e563a0ffb7d275eb3a889b\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"19d71f8c9a2056c0be4bb4fd4a75a325\"",
  "upstreamModelId": "openai-gpt-6-sol-xhigh-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 6906.642396,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-14

9.5 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the standard harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts.",
  "modelsEtag": "W/\"57cc4a7689e563a0ffb7d275eb3a889b\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"19d71f8c9a2056c0be4bb4fd4a75a325\"",
  "upstreamModelId": "openai-gpt-6-sol-none",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4655.549962,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-14

0.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the standard harness.",
  "modelsEtag": "W/\"57cc4a7689e563a0ffb7d275eb3a889b\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"19d71f8c9a2056c0be4bb4fd4a75a325\"",
  "upstreamModelId": "openai-gpt-6-sol-max",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5554.493172,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-14

4.6 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"57cc4a7689e563a0ffb7d275eb3a889b\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"19d71f8c9a2056c0be4bb4fd4a75a325\"",
  "upstreamModelId": "openai-gpt-6-sol-max-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 8721.949674,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-14

23.0 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"57cc4a7689e563a0ffb7d275eb3a889b\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"19d71f8c9a2056c0be4bb4fd4a75a325\"",
  "upstreamModelId": "openai-gpt-6-sol-none-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4233.770312,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-14

1.2 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-low-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'low' reasoning effort using the provider adapter harness.",
  "modelsEtag": "W/\"57cc4a7689e563a0ffb7d275eb3a889b\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"19d71f8c9a2056c0be4bb4fd4a75a325\"",
  "upstreamModelId": "openai-gpt-6-sol-low-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4745.286212,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-14

1.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"e7f673a6db195dac7f2956e3b1651145\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"4d3b46c6ffe9404ca1c0ebcfb5938883\"",
  "upstreamModelId": "openai-gpt-6-sol-none-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4233.770312,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-13

1.2 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-high-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"e7f673a6db195dac7f2956e3b1651145\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"4d3b46c6ffe9404ca1c0ebcfb5938883\"",
  "upstreamModelId": "openai-gpt-6-sol-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5908.428836,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-13

5.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (XHigh) · effort xhigh · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-xhigh-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'xhigh' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"e7f673a6db195dac7f2956e3b1651145\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"4d3b46c6ffe9404ca1c0ebcfb5938883\"",
  "upstreamModelId": "openai-gpt-6-sol-xhigh-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 6906.642396,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-13

9.5 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"e7f673a6db195dac7f2956e3b1651145\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"4d3b46c6ffe9404ca1c0ebcfb5938883\"",
  "upstreamModelId": "openai-gpt-6-sol-max-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 8721.949674,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-13

23.0 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-medium-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'medium' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"e7f673a6db195dac7f2956e3b1651145\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"4d3b46c6ffe9404ca1c0ebcfb5938883\"",
  "upstreamModelId": "openai-gpt-6-sol-medium-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5889.174868,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-13

4.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the standard harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts.",
  "modelsEtag": "W/\"e7f673a6db195dac7f2956e3b1651145\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"4d3b46c6ffe9404ca1c0ebcfb5938883\"",
  "upstreamModelId": "openai-gpt-6-sol-none",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4655.549962,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-13

0.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-low-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'low' reasoning effort using the provider adapter harness.",
  "modelsEtag": "W/\"e7f673a6db195dac7f2956e3b1651145\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"4d3b46c6ffe9404ca1c0ebcfb5938883\"",
  "upstreamModelId": "openai-gpt-6-sol-low-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4745.286212,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-13

1.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the standard harness.",
  "modelsEtag": "W/\"e7f673a6db195dac7f2956e3b1651145\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"4d3b46c6ffe9404ca1c0ebcfb5938883\"",
  "upstreamModelId": "openai-gpt-6-sol-max",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5554.493172,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-13

4.6 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the standard harness.",
  "modelsEtag": "W/\"b57a26ca4df83acaa40eadb6632c7c61\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"ee35c9dbbb033dc7de85467eedb85d54\"",
  "upstreamModelId": "openai-gpt-6-sol-max",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5554.493172,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-06

4.6 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"b57a26ca4df83acaa40eadb6632c7c61\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"ee35c9dbbb033dc7de85467eedb85d54\"",
  "upstreamModelId": "openai-gpt-6-sol-max-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 8721.949674,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-06

23.0 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"b57a26ca4df83acaa40eadb6632c7c61\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"ee35c9dbbb033dc7de85467eedb85d54\"",
  "upstreamModelId": "openai-gpt-6-sol-none-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4233.770312,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-06

1.2 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-low-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'low' reasoning effort using the provider adapter harness.",
  "modelsEtag": "W/\"b57a26ca4df83acaa40eadb6632c7c61\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"ee35c9dbbb033dc7de85467eedb85d54\"",
  "upstreamModelId": "openai-gpt-6-sol-low-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4745.286212,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-06

1.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-medium-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'medium' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"b57a26ca4df83acaa40eadb6632c7c61\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"ee35c9dbbb033dc7de85467eedb85d54\"",
  "upstreamModelId": "openai-gpt-6-sol-medium-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5889.174868,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-06

4.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-high-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"b57a26ca4df83acaa40eadb6632c7c61\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"ee35c9dbbb033dc7de85467eedb85d54\"",
  "upstreamModelId": "openai-gpt-6-sol-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5908.428836,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-06

5.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (XHigh) · effort xhigh · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-xhigh-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'xhigh' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"b57a26ca4df83acaa40eadb6632c7c61\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"ee35c9dbbb033dc7de85467eedb85d54\"",
  "upstreamModelId": "openai-gpt-6-sol-xhigh-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 6906.642396,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-06

9.5 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the standard harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts.",
  "modelsEtag": "W/\"b57a26ca4df83acaa40eadb6632c7c61\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"ee35c9dbbb033dc7de85467eedb85d54\"",
  "upstreamModelId": "openai-gpt-6-sol-none",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4655.549962,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-06

0.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (XHigh) · effort xhigh · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-xhigh-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'xhigh' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"b52a6e82589fb2c719168d69aa0bb015\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"f187ceaff4d7f6156e77c1f1254aafc8\"",
  "upstreamModelId": "openai-gpt-6-sol-xhigh-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 6906.642396,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-02

9.5 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-high-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"b52a6e82589fb2c719168d69aa0bb015\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"f187ceaff4d7f6156e77c1f1254aafc8\"",
  "upstreamModelId": "openai-gpt-6-sol-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5908.428836,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-02

5.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"b52a6e82589fb2c719168d69aa0bb015\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"f187ceaff4d7f6156e77c1f1254aafc8\"",
  "upstreamModelId": "openai-gpt-6-sol-none-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4233.770312,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-02

1.2 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-medium-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'medium' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"b52a6e82589fb2c719168d69aa0bb015\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"f187ceaff4d7f6156e77c1f1254aafc8\"",
  "upstreamModelId": "openai-gpt-6-sol-medium-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5889.174868,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-02

4.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the standard harness.",
  "modelsEtag": "W/\"b52a6e82589fb2c719168d69aa0bb015\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"f187ceaff4d7f6156e77c1f1254aafc8\"",
  "upstreamModelId": "openai-gpt-6-sol-max",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5554.493172,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-02

4.6 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the standard harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts.",
  "modelsEtag": "W/\"b52a6e82589fb2c719168d69aa0bb015\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"f187ceaff4d7f6156e77c1f1254aafc8\"",
  "upstreamModelId": "openai-gpt-6-sol-none",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4655.549962,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-02

0.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-low-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'low' reasoning effort using the provider adapter harness.",
  "modelsEtag": "W/\"b52a6e82589fb2c719168d69aa0bb015\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"f187ceaff4d7f6156e77c1f1254aafc8\"",
  "upstreamModelId": "openai-gpt-6-sol-low-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4745.286212,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-02

1.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"b52a6e82589fb2c719168d69aa0bb015\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"f187ceaff4d7f6156e77c1f1254aafc8\"",
  "upstreamModelId": "openai-gpt-6-sol-max-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 8721.949674,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-02

23.0 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"b52a6e82589fb2c719168d69aa0bb015\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"f187ceaff4d7f6156e77c1f1254aafc8\"",
  "upstreamModelId": "openai-gpt-6-sol-none-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4233.770312,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-01

1.2 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the standard harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts.",
  "modelsEtag": "W/\"b52a6e82589fb2c719168d69aa0bb015\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"f187ceaff4d7f6156e77c1f1254aafc8\"",
  "upstreamModelId": "openai-gpt-6-sol-none",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4655.549962,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-01

0.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the standard harness.",
  "modelsEtag": "W/\"b52a6e82589fb2c719168d69aa0bb015\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"f187ceaff4d7f6156e77c1f1254aafc8\"",
  "upstreamModelId": "openai-gpt-6-sol-max",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5554.493172,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-01

4.6 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-low-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'low' reasoning effort using the provider adapter harness.",
  "modelsEtag": "W/\"b52a6e82589fb2c719168d69aa0bb015\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"f187ceaff4d7f6156e77c1f1254aafc8\"",
  "upstreamModelId": "openai-gpt-6-sol-low-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4745.286212,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-01

1.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-medium-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'medium' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"b52a6e82589fb2c719168d69aa0bb015\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"f187ceaff4d7f6156e77c1f1254aafc8\"",
  "upstreamModelId": "openai-gpt-6-sol-medium-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5889.174868,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-01

4.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-high-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"b52a6e82589fb2c719168d69aa0bb015\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"f187ceaff4d7f6156e77c1f1254aafc8\"",
  "upstreamModelId": "openai-gpt-6-sol-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5908.428836,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-01

5.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (XHigh) · effort xhigh · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-xhigh-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'xhigh' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"b52a6e82589fb2c719168d69aa0bb015\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"f187ceaff4d7f6156e77c1f1254aafc8\"",
  "upstreamModelId": "openai-gpt-6-sol-xhigh-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 6906.642396,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-01

9.5 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"b52a6e82589fb2c719168d69aa0bb015\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"f187ceaff4d7f6156e77c1f1254aafc8\"",
  "upstreamModelId": "openai-gpt-6-sol-max-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 8721.949674,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-01

23.0 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the standard harness.",
  "modelsEtag": "W/\"8bfe2e142f29acec39df406a6631f81a\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"92bfaa47028eae49163960f4c5a298a0\"",
  "upstreamModelId": "openai-gpt-6-sol-max",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5554.493172,
  "providerModelIdPublished": false
}

Evidence observed 2026-09-30

4.6 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (XHigh) · effort xhigh · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-xhigh-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'xhigh' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"8bfe2e142f29acec39df406a6631f81a\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"92bfaa47028eae49163960f4c5a298a0\"",
  "upstreamModelId": "openai-gpt-6-sol-xhigh-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 6906.642396,
  "providerModelIdPublished": false
}

Evidence observed 2026-09-30

9.5 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-high-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"8bfe2e142f29acec39df406a6631f81a\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"92bfaa47028eae49163960f4c5a298a0\"",
  "upstreamModelId": "openai-gpt-6-sol-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5908.428836,
  "providerModelIdPublished": false
}

Evidence observed 2026-09-30

5.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-medium-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'medium' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"8bfe2e142f29acec39df406a6631f81a\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"92bfaa47028eae49163960f4c5a298a0\"",
  "upstreamModelId": "openai-gpt-6-sol-medium-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5889.174868,
  "providerModelIdPublished": false
}

Evidence observed 2026-09-30

4.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"8bfe2e142f29acec39df406a6631f81a\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"92bfaa47028eae49163960f4c5a298a0\"",
  "upstreamModelId": "openai-gpt-6-sol-none-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4233.770312,
  "providerModelIdPublished": false
}

Evidence observed 2026-09-30

1.2 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-low-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'low' reasoning effort using the provider adapter harness.",
  "modelsEtag": "W/\"8bfe2e142f29acec39df406a6631f81a\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"92bfaa47028eae49163960f4c5a298a0\"",
  "upstreamModelId": "openai-gpt-6-sol-low-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4745.286212,
  "providerModelIdPublished": false
}

Evidence observed 2026-09-30

1.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"8bfe2e142f29acec39df406a6631f81a\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"92bfaa47028eae49163960f4c5a298a0\"",
  "upstreamModelId": "openai-gpt-6-sol-max-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 8721.949674,
  "providerModelIdPublished": false
}

Evidence observed 2026-09-30

23.0 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the standard harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts.",
  "modelsEtag": "W/\"8bfe2e142f29acec39df406a6631f81a\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"92bfaa47028eae49163960f4c5a298a0\"",
  "upstreamModelId": "openai-gpt-6-sol-none",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4655.549962,
  "providerModelIdPublished": false
}

Evidence observed 2026-09-30

0.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the standard harness.",
  "modelsEtag": "W/\"16793624264c3bdded076b4e26af5a90\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"e7fb37556a7353e7da435c4c7b78ca50\"",
  "upstreamModelId": "openai-gpt-6-sol-max",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5554.493172,
  "providerModelIdPublished": false
}

Evidence observed 2026-09-29

4.6 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-high-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"16793624264c3bdded076b4e26af5a90\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"e7fb37556a7353e7da435c4c7b78ca50\"",
  "upstreamModelId": "openai-gpt-6-sol-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5908.428836,
  "providerModelIdPublished": false
}

Evidence observed 2026-09-29

5.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"16793624264c3bdded076b4e26af5a90\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"e7fb37556a7353e7da435c4c7b78ca50\"",
  "upstreamModelId": "openai-gpt-6-sol-none-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4233.770312,
  "providerModelIdPublished": false
}

Evidence observed 2026-09-29

1.2 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the standard harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts.",
  "modelsEtag": "W/\"16793624264c3bdded076b4e26af5a90\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"e7fb37556a7353e7da435c4c7b78ca50\"",
  "upstreamModelId": "openai-gpt-6-sol-none",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4655.549962,
  "providerModelIdPublished": false
}

Evidence observed 2026-09-29

0.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-low-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'low' reasoning effort using the provider adapter harness.",
  "modelsEtag": "W/\"16793624264c3bdded076b4e26af5a90\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"e7fb37556a7353e7da435c4c7b78ca50\"",
  "upstreamModelId": "openai-gpt-6-sol-low-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4745.286212,
  "providerModelIdPublished": false
}

Evidence observed 2026-09-29

1.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-medium-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'medium' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"16793624264c3bdded076b4e26af5a90\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"e7fb37556a7353e7da435c4c7b78ca50\"",
  "upstreamModelId": "openai-gpt-6-sol-medium-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5889.174868,
  "providerModelIdPublished": false
}

Evidence observed 2026-09-29

4.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (XHigh) · effort xhigh · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-xhigh-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'xhigh' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"16793624264c3bdded076b4e26af5a90\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"e7fb37556a7353e7da435c4c7b78ca50\"",
  "upstreamModelId": "openai-gpt-6-sol-xhigh-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 6906.642396,
  "providerModelIdPublished": false
}

Evidence observed 2026-09-29

9.5 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"16793624264c3bdded076b4e26af5a90\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"e7fb37556a7353e7da435c4c7b78ca50\"",
  "upstreamModelId": "openai-gpt-6-sol-max-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 8721.949674,
  "providerModelIdPublished": false
}

Evidence observed 2026-09-29

23.0 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (XHigh) · effort xhigh · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-xhigh-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'xhigh' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"284ac3aff89a7d8a3aca179ee1e5dd61\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"6e0849a05a2c572b1050e5900a372e2b\"",
  "upstreamModelId": "openai-gpt-6-sol-xhigh-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 6906.642396,
  "providerModelIdPublished": false
}

Evidence observed 2026-09-28

9.5 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the standard harness.",
  "modelsEtag": "W/\"284ac3aff89a7d8a3aca179ee1e5dd61\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"6e0849a05a2c572b1050e5900a372e2b\"",
  "upstreamModelId": "openai-gpt-6-sol-max",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5554.493172,
  "providerModelIdPublished": false
}

Evidence observed 2026-09-28

4.6 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Max) · effort max · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-max-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'max' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"284ac3aff89a7d8a3aca179ee1e5dd61\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"6e0849a05a2c572b1050e5900a372e2b\"",
  "upstreamModelId": "openai-gpt-6-sol-max-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 8721.949674,
  "providerModelIdPublished": false
}

Evidence observed 2026-09-28

23.0 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the provider adapter harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"284ac3aff89a7d8a3aca179ee1e5dd61\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"6e0849a05a2c572b1050e5900a372e2b\"",
  "upstreamModelId": "openai-gpt-6-sol-none-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4233.770312,
  "providerModelIdPublished": false
}

Evidence observed 2026-09-28

1.2 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-low-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'low' reasoning effort using the provider adapter harness.",
  "modelsEtag": "W/\"284ac3aff89a7d8a3aca179ee1e5dd61\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"6e0849a05a2c572b1050e5900a372e2b\"",
  "upstreamModelId": "openai-gpt-6-sol-low-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4745.286212,
  "providerModelIdPublished": false
}

Evidence observed 2026-09-28

1.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-medium-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'medium' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"284ac3aff89a7d8a3aca179ee1e5dd61\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"6e0849a05a2c572b1050e5900a372e2b\"",
  "upstreamModelId": "openai-gpt-6-sol-medium-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5889.174868,
  "providerModelIdPublished": false
}

Evidence observed 2026-09-28

4.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-high-provider-adapter

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter harness. Some model actions have unmeasured token usage; reported costs are recorded-usage lower bounds, not verified full-run costs.",
  "modelsEtag": "W/\"284ac3aff89a7d8a3aca179ee1e5dd61\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"6e0849a05a2c572b1050e5900a372e2b\"",
  "upstreamModelId": "openai-gpt-6-sol-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 5908.428836,
  "providerModelIdPublished": false
}

Evidence observed 2026-09-28

5.9 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · GPT-6 Sol (None) · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:openai-gpt-6-sol-none

View run source

Source metadata
{
  "provider": "OpenAI",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "openai-gpt-6-sol",
  "modelNotes": "Evaluation conducted with 'none' reasoning effort using the standard harness. Partial dataset coverage: scores use full catalog denominators; costs include recorded usage only, excluding unrecorded attempts.",
  "modelsEtag": "W/\"284ac3aff89a7d8a3aca179ee1e5dd61\"",
  "scoreMetric": "relative-human-action-efficiency",
  "evaluationsUrl": "https://arcprize.org/media/data/evaluations.json",
  "evaluationsEtag": "W/\"6e0849a05a2c572b1050e5900a372e2b\"",
  "upstreamModelId": "openai-gpt-6-sol-none",
  "environmentCount": 25,
  "modelReleaseDate": "2026-09-22T00:00:00.000Z",
  "upstreamDatasetId": "v3_Semi_Private",
  "actionBudgetPolicy": "5x-human-baseline-median-per-level",
  "technicalReportUrl": "https://arcprize.org/media/ARC_AGI_3_Technical_Report.pdf",
  "catalogMatchAliases": [],
  "agentHarnessPublished": false,
  "verificationAuthority": "ARC Prize Foundation",
  "upstreamRunIdPublished": false,
  "actualEvaluationCostUsd": 4655.549962,
  "providerModelIdPublished": false
}

Evidence observed 2026-09-28

0.3 / 100

How GPT-6 Sol Compares

Axes

Filters

$0.00 – $67.5
≥ 1.4

Providers

ai2ai21-labsalibabaanthropicarceeawsazurebaidu
openai
zai
meta
nvidia
xai
upstage
deepseek
alibaba
mistral
nous-research
aws
google
anthropic
xiaomi
stepfun
kimi
reflection
minimax
azure
inclusionai
reka-ai
cohere
bytedance_seed
inception
tencent
thinking-machines
Bubble size = Context Window
8191.00
2000000.00
Compare with other models