- Wildbench v2 65.3
- Arena Hard v2 43.1
- IF (Internal; accuracy) 84.8
- Infinite Generations (Internal; lower is better) 1.29
- MMLU 80.5
- MMLU Pro (5-shot CoT) 69.1
- MATH 69.4
- GPQA Main (5-shot CoT) 44.2
- GPQA Diamond (5-shot CoT) 46.1
- MBPP Plus - Pass@5 78.3
- HumanEval Plus - Pass@5 92.9
- SimpleQA (TotalAcc) 12.1
- MMMU 62.5
- Mathvista 67.1
- ChartQA 87.4
- DocVQA 94.9
- AI2D 92.9
Mistral Small 3.2
Mistral Small 3.2 places in the top three on GPQA Main.
38
Benchmarks with a score
0
Ranked first, out of five or more models
1
In the top three
Mistral AI Jun 2025 Mistral Small 3.2
Open weights
123.7
#324 of 404
Self-Reported Index, from the lab's own numbers. No Epoch Capabilities Index score yet.
- GPQA Diamond 49% #158 /229
- Human Pathogen Capabilities Test v2 (HPCT v2) 27.9% #66 /71
- Phare 64.5% #35 /62
Where it ranks
Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.
Knowledge and language
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Slop Score | 52.7 | #8 of 17 | Official leaderboard | |
| LMArena (Elo) | 1,356 | #171 of 335 | Official leaderboard | |
| Phare | 64.5% | #35 of 62 | Official leaderboard | |
| MMLU | 80.5% | #74 of 131 | Lab model card | |
| Wer wird Millionär? LLM Benchmark | 933.0 | #49 of 77 | Official leaderboard | |
| EQ-Bench 3 | 1,091 | #56 of 78 | Official leaderboard | |
| MMLU-Pro | 69.1% | #124 of 167 | Lab model card | |
| Creative Writing v3 | 57.2 | #90 of 120 | Official leaderboard | |
| SimpleQA | 12.1% | #47 of 60 | Lab model card | |
| Longform Creative Writing (EQ-Bench) | 41.6 | #97 of 119 | Official leaderboard | |
| Judgemark v2.1 | 50.2 | #39 of 42 | Official leaderboard | |
| Arena-Hard v2.0 | 43.1% | #24 of 24 | Lab model card | |
| IFEval (Mistral internal version) | 84.8% | #2 of 2 | Lab model card | |
| Infinite Generations (Mistral internal) | 1.29% | #1 of 1 | Lab model card |
Math and science
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| GPQA Main | 44.2% | #3 of 7 | Lab model card | |
| MATH | 69.4% | #47 of 99 | Lab model card | |
| OTIS Mock AIME 2024-2025 | 30.3% | #146 of 212 | Epoch AI run | |
| GPQA Diamond | 49% | #158 of 229 | Epoch AI run | |
| DTBench | 59.9% | #145 of 180 | Official leaderboard | |
| Verifying the Verifiers (Project APE CRED benchmark) | 19% | #21 of 23 | Official leaderboard |
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| HumanEval+ | 92.9% | #4 of 27 | Lab model card | |
| MBPP+ | 78.3% | #7 of 15 | Lab model card | |
| DesignArena Website | 889 | #142 of 151 | Official leaderboard | |
| DesignArena UI Component | 898 | #133 of 139 | Official leaderboard | |
| DesignArena Code | 897 | #139 of 143 | Official leaderboard | |
| DesignArena Game Dev | 878 | #140 of 143 | Official leaderboard |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| BFCL V4 | 37.1% | #32 of 65 | Official leaderboard | |
| Galileo Agent Leaderboard v2 | 26% | #18 of 21 | Official leaderboard | |
| LisanBench | 52.7 | #122 of 133 | Official leaderboard |
Vision, audio and robotics
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| AI2D | 92.9% | #13 of 52 | Lab model card | |
| DocVQA | 94.9% | #13 of 36 | Lab model card | |
| ChartQA | 87.4% | #15 of 37 | Lab model card | |
| MathVista | 67.1% | #54 of 70 | Lab model card | |
| Vision Arena | 1,140 | #105 of 136 | Official leaderboard | |
| MMMU | 62.5% | #81 of 103 | Lab model card | |
| MeasureBench Real-world | 8.50% | #18 of 19 | Official leaderboard | |
| MeasureBench Synthetic | 6.50% | #18 of 19 | Official leaderboard |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Human Pathogen Capabilities Test v2 (HPCT v2) | 27.9% | #66 of 71 | Official leaderboard |
Its lab's numbers put it at 123.7, number 324 of 404
The Self-Reported Index fits one score to the 13 capability results Mistral AI reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 121.3 to 126.1. How we compute it.
- DocVQA 94.9
- AI2D 92.9
- HumanEval+ 92.9
- ChartQA 87.4
- MMLU 80.5
- MBPP+ 78.3
- MATH 69.4
- MMLU-Pro 69.1
- MathVista 67.1
- MMMU 62.5
- GPQA Diamond 46.1
- Arena-Hard v2.0 43.1
- SimpleQA 12.1
Its lab's number next to a leaderboard
Where the benchmark's own leaderboard, or Epoch AI's run, also has this model. Gaps mostly come from different harnesses, effort levels and task subsets. Every pair.
| Benchmark | Lab says | Board says | Gap |
|---|---|---|---|
| GPQA Diamond | 46.1% GPQA Diamond (5-shot CoT) | 49% Epoch AI run | −2.9 pts |
Its launch post
The benchmarks the lab chose to headline, with the scores it gave.