Ministral 3 3B

December 2, 2025Last 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

$0.10/M

Input Price

$0.10/M

Output Price

$0.10/M

Speed

181 tok/s

TTFT

0.34s

Benchmark Scores

Intelligence Index

Performance Radar

Ministral 3 3B

Field Average

External Benchmarks

AA-Omniscience Accuracy

How Ministral 3 3B 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

0.4 / 100

LiveCodeBench

0.2 / 100

Terminal-Bench Hard

0.0 / 100

τ²-Bench

0.2 / 100

AA-Omniscience Hallucination RatePredicted

97.9 / 100

Aider PolyglotPredicted

51.8 / 100

AIMEPredicted

0.0 / 30

AlpacaEval 2.0 LCPredicted

14.5 / 100

AlpacaEval 2.0 RawPredicted

11.8 / 100

ARC-AGI-1Predicted

92.8 / 100

ARC-AGI-1 Cost per TaskPredicted

2.4

ARC-AGI-2Predicted

84.4 / 100

ARC-AGI-2 Cost per TaskPredicted

4.2

BFCL (Berkeley Function Calling)Predicted

32.4

BigCodeBench CompletePredicted

24.0 / 100

BigCodeBench InstructPredicted

18.7 / 100

AA Intelligence Index (Matrix)Predicted

49.1

AA Long Context Reasoning (Matrix)Predicted

78.7

AIME 2024Predicted

84.2

AIME 2025 (Matrix)Predicted

99.3

Arena-Hard AutoPredicted

29.5

BrowseCompPredicted

89.8

BRUMO 2025Predicted

99.9

CMIMC 2025Predicted

96.8

CritPtPredicted

41.1

GPQA Diamond (Matrix)Predicted

70.7

HLE (Matrix)Predicted

49.6

HMMT Feb 2025Predicted

92.4

HMMT Nov 2025Predicted

95.3

HumanEvalPredicted

77.8

IFBench (Matrix)Predicted

54.0

IFEvalPredicted

83.7

IMO 2025Predicted

62.7

LiveCodeBench (Matrix)Predicted

57.9

MATH-500 (Matrix)Predicted

93.7

MathArena Apex 2025Predicted

32.1

MMLU-Pro (Matrix)Predicted

51.1

MMMU-ProPredicted

76.5

MRCR v2Predicted

83.6

OSWorldPredicted

82.2

SimpleQAPredicted

66.9

SMT 2025Predicted

95.9

SWE-bench ProPredicted

61.8

Tau-Bench Telecom (Matrix)Predicted

99.3

Terminal-Bench 2.0Predicted

85.2

Terminal-Bench 1.0Predicted

47.5

USAMO 2025Predicted

23.3

Video-MMUPredicted

84.1

browsecompPredicted

91.4

BullshitBenchPredicted

61.1 / 100

Aider PolyglotPredicted

0.1

Apex AgentsPredicted

4.9

Arc Agi 2Predicted

1.6

BIG-Bench HardPredicted

3.0

CAD-EvalPredicted

5.0

Chess PuzzlesPredicted

0.5

CyBenchPredicted

0.4

DeepResearchBenchPredicted

0.5

FictionLiveBenchPredicted

0.6

GeoBenchPredicted

0.0

GSM8K (Epoch)Predicted

5.3

HellaSwagPredicted

0.0

Lech Mazur WritingPredicted

7.8

METR Time HorizonsPredicted

42.5

OTIS Mock AIME 2024–2025Predicted

0.1

PosttrainbenchPredicted

0.0

SimpleQA Verified (Epoch)Predicted

0.8

The Agent CompanyPredicted

1.4

TriviaQAPredicted

21.1

WinoGrandePredicted

0.7

FrontierMathPredicted

53.8 / 100

GAIA Level 1Predicted

21.0

GAIA Level 2Predicted

2.5

GAIA Level 3Predicted

0.0

GAIAPredicted

12.4 / 100

LegalBenchPredicted

24.1 / 100

LiveBench CodingPredicted

81.3 / 100

LiveBench Data AnalysisPredicted

77.2 / 100

LiveBench LanguagePredicted

87.0 / 100

LiveBench MathPredicted

91.1 / 100

LiveBench OverallPredicted

81.4 / 100

LiveBench ReasoningPredicted

87.2 / 100

LongBench v2 EasyPredicted

35.8

LongBench v2 HardPredicted

27.3

LongBench v2Predicted

25.5 / 100

MATH-500Predicted

0.6 / 100

MathVistaPredicted

19.5 / 100

MedQA (USMLE)Predicted

74.4

MLE-benchPredicted

78.6 / 100

MMMUPredicted

67.9 / 100

MMTU Table UnderstandingPredicted

59.2 / 100

MT-BenchPredicted

7.7 / 10

NoLiMa (NIAH)Predicted

84.3 / 100

OCRBench v2Predicted

59.6 / 100

Open LLM AveragePredicted

10.7 / 100

Open LLM: BBHPredicted

32.2 / 100

Open LLM: GPQAPredicted

24.6 / 100

Open LLM: IFEvalPredicted

37.6 / 100

Open LLM: MATH Level 5Predicted

2.8 / 100

Open LLM: MMLU-PROPredicted

18.2 / 100

Open LLM: MUSRPredicted

35.2 / 100

RE-BenchPredicted

100.0

SimpleBenchPredicted

62.4 / 100

simpleqaPredicted

54.9

SWE-bench LitePredicted

22.4 / 100

SWE-bench VerifiedPredicted

69.5 / 100

tau-bench RetailPredicted

94.3 / 100

WebArenaPredicted

0.0 / 100

WeirdMLPredicted

39.8 / 100

WildBenchPredicted

35.3

Vectara Factual Consistencyvectara_hallucination

75.8 / 100

Vectara Hallucination Ratevectara_hallucination

24.2 / 100