GPT-5.4 mini (xhigh)

March 17, 2026Last synced May 28, 2026, 9:09 AM

Human Frontier

—

Human-calibrated frontier signal, backed by Arena-style preference evidence and separate from raw AgMoBench benchmark composite scores.

Blended Price

$1.69/M

Input Price

$0.75/M

Output Price

$4.50/M

Speed

153 tok/s

TTFT

7.19s

Benchmark Scores

Intelligence Index

Performance Radar

GPT-5.4 mini (xhigh)

Field Average

External Benchmarks

AA-Omniscience Hallucination Rate

How GPT-5.4 mini (xhigh) Compares

Axes

X Axis

Y Axis

Bubble Size

Filters

Blended Price (USD)$0.00 – $37.5

AgMoBench Overall≥ 2.1

Providers

ai2ai21-labsalibabaanthropicarceeawsazurebaidu

Show quadrants

mistral

meta

openai

google

anthropic

azure

nvidia

alibaba

aws

ibm

nous-research

kimi

xai

zai

deepcogito

kwaikat

xiaomi

cohere

ai21-labs

inclusionai

minimax

baidu

deepseek

prime-intellect

liquidai

stepfun

bytedance_seed

upstage

reka-ai

inception

tencent

arcee

ai2

swiss-ai-initiative

perplexity

Bubble size = Context Window

8191.00

2000000.00

Compare with other models

GPQA Diamond

0.9 / 100

LiveCodeBench

— / 100

Terminal-Bench Hard

0.5 / 100

τ²-Bench

0.8 / 100

AA-Omniscience Indexaa_omniscience

-18.7 / 100

ARC-AGI-1arc_agi

13.0 / 100

ARC-AGI-1 Cost per Taskarc_agi

0.0

ARC-AGI-2arc_agi

1.1 / 100

ARC-AGI-2 Cost per Taskarc_agi

0.1

AA-Omniscience AccuracyPredicted

41.6 / 100

Aider PolyglotPredicted

83.8 / 100

AIMEPredicted

0.8 / 30

AIME 2025Predicted

0.9 / 30

AlpacaEval 2.0 LCPredicted

30.4 / 100

AlpacaEval 2.0 RawPredicted

23.6 / 100

BFCL (Berkeley Function Calling)Predicted

58.3

BigCodeBench CompletePredicted

53.6 / 100

BigCodeBench InstructPredicted

43.4 / 100

AA Intelligence Index (Matrix)Predicted

66.9

AA Long Context Reasoning (Matrix)Predicted

73.4

AIME 2024Predicted

98.8

AIME 2025 (Matrix)Predicted

99.5

Arena-Hard AutoPredicted

72.9

BrowseCompPredicted

65.4

BRUMO 2025Predicted

96.7

CMIMC 2025Predicted

86.0

GPQA Diamond (Matrix)Predicted

90.0

HLE (Matrix)Predicted

28.3

HMMT Feb 2025Predicted

90.5

HMMT Nov 2025Predicted

91.2

HumanEvalPredicted

94.9

IFBench (Matrix)Predicted

61.5

IFEvalPredicted

92.2

IMO 2025Predicted

28.5

LiveCodeBench (Matrix)Predicted

84.9

MATH-500 (Matrix)Predicted

98.6

MathArena Apex 2025Predicted

2.0

MMLU-Pro (Matrix)Predicted

87.4

MMMU-ProPredicted

82.1

MRCR v2Predicted

78.2

OSWorldPredicted

44.2

SimpleQAPredicted

46.4

SMT 2025Predicted

88.6

SWE-bench ProPredicted

41.5

Tau-Bench Telecom (Matrix)Predicted

97.3

Terminal-Bench 2.0Predicted

34.7

Terminal-Bench 1.0Predicted

33.7

USAMO 2025Predicted

22.0

Video-MMUPredicted

86.7

browsecompPredicted

62.5

Aider PolyglotPredicted

0.1

Apex AgentsPredicted

2.4

Arc Agi 2Predicted

0.0

BIG-Bench HardPredicted

3.0

CAD-EvalPredicted

2.6

Chess PuzzlesPredicted

0.2

CyBenchPredicted

0.6

DeepResearchBenchPredicted

0.4

FictionLiveBenchPredicted

0.8

GeoBenchPredicted

0.0

GSM8K (Epoch)Predicted

2.9

HellaSwagPredicted

0.0

Lech Mazur WritingPredicted

8.5

METR Time HorizonsPredicted

10.6

OTIS Mock AIME 2024–2025Predicted

1.3

PosttrainbenchPredicted

0.0

SimpleQA Verified (Epoch)Predicted

0.4

The Agent CompanyPredicted

15.3

TriviaQAPredicted

10.7

WinoGrandePredicted

0.9

FrontierMathPredicted

19.9 / 100

GAIA Level 1Predicted

86.1

GAIA Level 2Predicted

81.0

GAIA Level 3Predicted

94.7

GAIAPredicted

74.2 / 100

LegalBenchPredicted

89.0 / 100

LiveCodeBenchPredicted

0.8 / 100

LongBench v2 EasyPredicted

71.3

LongBench v2 HardPredicted

56.9

LongBench v2Predicted

45.2 / 100

MATH-500Predicted

1.0 / 100

MathVistaPredicted

74.1 / 100

MedQA (USMLE)Predicted

95.7

MLE-benchPredicted

42.7 / 100

MMLU ProPredicted

0.9 / 100

MMMUPredicted

84.0 / 100

MMTU Table UnderstandingPredicted

68.7 / 100

MT-BenchPredicted

7.9 / 10

NoLiMa (NIAH)Predicted

95.1 / 100

OCRBench v2Predicted

86.0 / 100

Open LLM AveragePredicted

20.6 / 100

Open LLM: BBHPredicted

51.0 / 100

Open LLM: GPQAPredicted

30.5 / 100

Open LLM: IFEvalPredicted

38.2 / 100

Open LLM: MATH Level 5Predicted

16.3 / 100

Open LLM: MMLU-PROPredicted

34.9 / 100

Open LLM: MUSRPredicted

41.8 / 100

RE-BenchPredicted

90.5

SimpleBenchPredicted

52.5 / 100

simpleqaPredicted

46.0

SWE-bench LitePredicted

56.5 / 100

SWE-bench VerifiedPredicted

74.7 / 100

tau-bench RetailPredicted

83.8 / 100

WebArenaPredicted

2.4 / 100

WildBenchPredicted

47.7

BullshitBenchbullshitbench

32.0 / 100

Apex Agentsepoch_ai

3.1

Arc Agi 2epoch_ai

0.2

Epoch Capabilities Indexepoch_ai

148.7

LiveBench Codinglivebench

71.6 / 100

LiveBench Data Analysislivebench

71.0 / 100

LiveBench Languagelivebench

71.5 / 100

LiveBench Mathlivebench

78.6 / 100

LiveBench Overalllivebench

67.5 / 100

LiveBench Reasoninglivebench

72.5 / 100

Vectara Factual Consistencyvectara_hallucination

94.5 / 100

Vectara Hallucination Ratevectara_hallucination

5.5 / 100

WeirdMLweirdml

60.3 / 100