Mistral Small 3.1.
Mistral AI Mar 2025 Open weights
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.
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| MMLU | 80.6% | #73 of 131 | Lab model card | |
| LMArena (Elo) | 1,303 | #231 of 335 | Official leaderboard | |
| Vectara Hallucination Leaderboard (HHEM) | 5.60% | #104 of 148 | Official leaderboard | |
| MMLU-Pro | 66.8% | #131 of 167 | Lab model card | |
| RULER | 81.2% | #33 of 42 | Lab model card | |
| Global MMLU Lite | 78.5% | #34 of 43 | Official leaderboard | |
| SimpleQA | 10.4% | #49 of 60 | Lab model card | |
| IFEval | 75% | #53 of 60 | Official leaderboard | |
| FACTS Parametric | 5.47% | #42 of 45 | Official leaderboard | |
| EQ-Bench 3 | 832 | #73 of 78 | Official leaderboard | |
| LongBench v2 | 37.2% | #34 of 35 | Lab model card | |
| Creative Writing v3 | 42.3 | #118 of 120 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| MATH | 69.3% | #48 of 99 | Lab model card | |
| MATH Level 5 | 46.8% | #54 of 94 | Epoch AI run | |
| GPQA Diamond | 47.5% | #166 of 229 | Epoch AI run | |
| GPQA Main | 39.2% | #46 of 60 | Official leaderboard | |
| OTIS Mock AIME 2024-2025 | 5.83% | #180 of 212 | Epoch AI run | |
| SciArena | 890 | #30 of 34 | Official leaderboard | |
| Misguided Attention | 26% | #41 of 45 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Galileo Agent Leaderboard | 0.91 | #6 of 29 | Official leaderboard | |
| Berkeley Function Calling Leaderboard (BFCL) | 57.7% | #31 of 67 | Official leaderboard | |
| MCP-Bench | 0.53 | #11 of 15 | Official leaderboard | |
| ITBench | 32.9% | #12 of 13 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| MortgageTax | 61.2% | #58 of 87 | Official leaderboard | |
| Vals Finance Agent v1 | 10% | #40 of 47 | Official leaderboard | |
| LegalBench | 69.2% | #118 of 138 | Official leaderboard | |
| FailureSensorIQ | 47.9% | #21 of 24 | Official leaderboard | |
| CorpFin v2 | 44.2% | #106 of 119 | Official leaderboard | |
| TaxEval v2 | 58.3% | #119 of 129 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| AI2D | 93.7% | #10 of 52 | Lab model card | |
| DocVQA | 94.1% | #18 of 36 | Lab model card | |
| ChartQA | 86.2% | #20 of 37 | Lab model card | |
| MathVista | 68.9% | #49 of 70 | Lab model card | |
| MMMU | 64% | #79 of 103 | Lab model card | |
| Vision Arena | 1,127 | #109 of 136 | Official leaderboard | |
| MM-MT-Bench | 7.30 | #5 of 6 | Lab model card | |
| MMMU-Pro | 49.3% | #80 of 90 | Lab model card |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| BaxBench | 10.8% | #25 of 34 | Official leaderboard | |
| Human Pathogen Capabilities Test | 23.2% | #83 of 86 | Official leaderboard |
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| HELM Capabilities | 55.8 | #49 of 60 | Official leaderboard | |
| Political Science LLM Benchmark | 64.2% | #35 of 42 | Official leaderboard |
The Self-Reported Index fits one score to the 15 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 121.6 to 125.4. How we compute it.
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.
The benchmarks the lab chose to headline, with the scores it gave.