AgMoDB
ModelsAgentsOrchestratorsEvalsCompositesVisualizeIndustry
AgMoDB by @mistakeknot
All Models

Gemini 4 Argon (High)

September 30, 2026Last synced Nov 22, 2026, 8:17 AM1049K context

Blended Price

$8.00/M

Input Price

$4.00/M

Output Price

$20.00/M

Speed

85 tok/s

TTFT

32.94s

Benchmark Scores

Intelligence Index
50.4
Coding Index
—
Math Index
—
MMLU Pro
— / 100
GPQA Diamond
— / 100
HLE
57.1 / 100
LiveCodeBench
— / 100
SciCode
61.8 / 100
MATH-500
— / 100
AIME
— / 30
AIME 2025
— / 30
IFBench
— / 100
LCR
79.7 / 100
Terminal-Bench Hard
— / 100
τ²-Bench
— / 100

Performance Radar

Gemini 4 Argon (High)
Field Average

Hidden, no data: AgMoBench Agentic

External Benchmarks

AA-Omniscience Accuracyaa_omniscience
49.9 / 100
AA-Omniscience Hallucination Rateaa_omniscience
15.1 / 100
AA-Omniscience Indexaa_omniscience
42.4 / 100
ARC-AGI-1arc_agi
76.5 / 100
ARC-AGI-1 Cost per Taskarc_agi
0.0
ARC-AGI-2arc_agi
42.1 / 100
ARC-AGI-2 Cost per Taskarc_agi
0.1
BullshitBenchbullshitbench
66.0 / 100
Arena ELO: Codingchatbot_arena
1682.0
Chatbot Arena ELOchatbot_arena
1528.0
Ale Benchepoch_ai
2100.1
Arc Agi 2epoch_ai
0.9
Chess Puzzlesepoch_ai
0.6
Critptepoch_ai
0.3
Cursorbenchepoch_ai
0.6
Dtbenchepoch_ai
1.0
Ebr Benchepoch_ai
0.6
Frontiermath Erdosepoch_ai
0.0
Frontiermath Tier 4 V2epoch_ai
0.8
Frontiermath Tiers 1 3 V2epoch_ai
0.9
Frontiersweepoch_ai
0.5
Furniture Assemblyepoch_ai
0.7
Gdp Pdfepoch_ai
0.5
Lmcaepoch_ai
63.7
Mirrorcodeepoch_ai
0.6
Mystery Game Puzzlesepoch_ai
0.6
OTIS Mock AIME 2024–2025epoch_ai
1.0
Posttrainbenchepoch_ai
0.4
Scicodeepoch_ai
0.6
SimpleQA Verified (Epoch)epoch_ai
0.8
WebDev Arenaepoch_ai
1679.5
Weirdml V3epoch_ai
0.2
LegalBenchlegalbench
88.3 / 100
LiveBench Codinglivebench
85.7 / 100
LiveBench Data Analysislivebench
85.2 / 100
LiveBench Reasoninglivebench
95.1 / 100
MCP Atlasseal
90.7
MultiChallengeseal
78.7
Vectara Factual Consistencyvectara_hallucination
87.9 / 100
Vectara Hallucination Ratevectara_hallucination
12.1 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-low

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon",
  "modelNotes": "Evaluation conducted with 'low' 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": "google-gemini-4-argon-low",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 5820.531006,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-20

49.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-medium

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon",
  "modelNotes": "Evaluation conducted with 'medium' 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": "google-gemini-4-argon-medium",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 6179.253994,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-20

46.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-high-provider-adapter

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter 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": "google-gemini-4-argon-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 4625.84128,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-20

97.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-low

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon",
  "modelNotes": "Evaluation conducted with 'low' 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": "google-gemini-4-argon-low",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 5820.531006,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-20

49.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-high-provider-adapter

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter 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": "google-gemini-4-argon-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 4625.84128,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-20

97.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-medium

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon",
  "modelNotes": "Evaluation conducted with 'medium' 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": "google-gemini-4-argon-medium",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 6179.253994,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-20

46.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-medium

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon",
  "modelNotes": "Evaluation conducted with 'medium' 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": "google-gemini-4-argon-medium",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 6179.253994,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-18

46.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-low

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon",
  "modelNotes": "Evaluation conducted with 'low' 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": "google-gemini-4-argon-low",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 5820.531006,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-18

49.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-high-provider-adapter

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter 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": "google-gemini-4-argon-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 4625.84128,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-18

97.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-low

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon",
  "modelNotes": "Evaluation conducted with 'low' 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": "google-gemini-4-argon-low",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 5820.531006,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-17

49.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-high-provider-adapter

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter 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": "google-gemini-4-argon-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 4625.84128,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-17

97.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-medium

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon",
  "modelNotes": "Evaluation conducted with 'medium' 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": "google-gemini-4-argon-medium",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 6179.253994,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-17

46.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-high-provider-adapter

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter 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": "google-gemini-4-argon-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 4625.84128,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-16

97.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-low

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon",
  "modelNotes": "Evaluation conducted with 'low' 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": "google-gemini-4-argon-low",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 5820.531006,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-16

49.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-medium

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon",
  "modelNotes": "Evaluation conducted with 'medium' 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": "google-gemini-4-argon-medium",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 6179.253994,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-16

46.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-medium

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon",
  "modelNotes": "Evaluation conducted with 'medium' 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": "google-gemini-4-argon-medium",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 6179.253994,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-10

46.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-low

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon",
  "modelNotes": "Evaluation conducted with 'low' 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": "google-gemini-4-argon-low",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 5820.531006,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-10

49.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-high-provider-adapter

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter 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": "google-gemini-4-argon-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 4625.84128,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-10

97.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-medium

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon",
  "modelNotes": "Evaluation conducted with 'medium' 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": "google-gemini-4-argon-medium",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 6179.253994,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-09

46.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-low

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon",
  "modelNotes": "Evaluation conducted with 'low' 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": "google-gemini-4-argon-low",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 5820.531006,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-09

49.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-high-provider-adapter

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter 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": "google-gemini-4-argon-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 4625.84128,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-09

97.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-high-provider-adapter

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter 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": "google-gemini-4-argon-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 4625.84128,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-06

97.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-medium

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon",
  "modelNotes": "Evaluation conducted with 'medium' 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": "google-gemini-4-argon-medium",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 6179.253994,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-06

46.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-low

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon",
  "modelNotes": "Evaluation conducted with 'low' 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": "google-gemini-4-argon-low",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 5820.531006,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-06

49.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-low

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon",
  "modelNotes": "Evaluation conducted with 'low' 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": "google-gemini-4-argon-low",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 5820.531006,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-04

49.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-medium

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon",
  "modelNotes": "Evaluation conducted with 'medium' 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": "google-gemini-4-argon-medium",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 6179.253994,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-04

46.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-high-provider-adapter

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter 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": "google-gemini-4-argon-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 4625.84128,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-04

97.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-high-provider-adapter

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter 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": "google-gemini-4-argon-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 4625.84128,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-02

97.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-low

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon",
  "modelNotes": "Evaluation conducted with 'low' 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": "google-gemini-4-argon-low",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 5820.531006,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-02

49.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-medium

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon",
  "modelNotes": "Evaluation conducted with 'medium' 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": "google-gemini-4-argon-medium",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 6179.253994,
  "providerModelIdPublished": false
}

Evidence observed 2026-11-02

46.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-medium

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon",
  "modelNotes": "Evaluation conducted with 'medium' 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": "google-gemini-4-argon-medium",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 6179.253994,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-28

46.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-high-provider-adapter

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter 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": "google-gemini-4-argon-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 4625.84128,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-28

97.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-low

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon",
  "modelNotes": "Evaluation conducted with 'low' 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": "google-gemini-4-argon-low",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 5820.531006,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-28

49.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-high-provider-adapter

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter 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": "google-gemini-4-argon-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 4625.84128,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-21

97.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-medium

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon",
  "modelNotes": "Evaluation conducted with 'medium' 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": "google-gemini-4-argon-medium",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 6179.253994,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-21

46.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-low

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon",
  "modelNotes": "Evaluation conducted with 'low' 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": "google-gemini-4-argon-low",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 5820.531006,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-21

49.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-medium

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon",
  "modelNotes": "Evaluation conducted with 'medium' 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": "google-gemini-4-argon-medium",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 6179.253994,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-19

46.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-high-provider-adapter

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter 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": "google-gemini-4-argon-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 4625.84128,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-19

97.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-low

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon",
  "modelNotes": "Evaluation conducted with 'low' 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": "google-gemini-4-argon-low",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 5820.531006,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-19

49.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon (Medium) · effort medium · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-medium

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon",
  "modelNotes": "Evaluation conducted with 'medium' 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": "google-gemini-4-argon-medium",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 6179.253994,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-15

46.3 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon (Low) · effort low · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-low

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon",
  "modelNotes": "Evaluation conducted with 'low' 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": "google-gemini-4-argon-low",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 5820.531006,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-15

49.8 / 100
ARC-AGI-3 Public Demoarc_agi

v3 · public-demo · Gemini 4 Argon - Provider Adapter (High) · effort high · 1 attempt · 5x-human-baseline-median-per-level · arc-agi-3:public-demo · record v3_Semi_Private:google-gemini-4-argon-high-provider-adapter

View run source

Source metadata
{
  "provider": "Google",
  "modelType": "CoT",
  "modelsUrl": "https://arcprize.org/media/data/models.json",
  "modelGroup": "google-gemini-4-argon-provider-adapter",
  "modelNotes": "Evaluation conducted with 'high' reasoning effort using the provider adapter 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": "google-gemini-4-argon-high-provider-adapter",
  "environmentCount": 25,
  "modelReleaseDate": "2026-10-13T00:00:00.000Z",
  "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": 4625.84128,
  "providerModelIdPublished": false
}

Evidence observed 2026-10-15

97.3 / 100

How Gemini 4 Argon (High) 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