Remote agents in Vibe, powered by Mistral Medium 3.5 May 22, 2026
Mistral Medium 3.5
Mistral Medium 3.5.
63
Benchmarks with a score
0
Ranked first, out of five or more models
0
In the top three
Mistral AI Apr 2026 Mistral Medium 3.5
Open weights Mid-size
141.6
#114 of 256
Epoch Capabilities Index
Self-Reported Index 143.5 #149 of 404
- IFBench 69% #46 /80
- Harvey LAB (Legal Agent Benchmark) 0.42% #61 /80
- Vals Finance Agent v2 32.1% #70 /80
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 | |
|---|---|---|---|---|
| LMArena (Elo) | 1,427 | #94 of 335 | Official leaderboard | |
| COLLIE | 95.8% | #4 of 12 | Lab model card | |
| IFBench | 69% | #46 of 80 | Lab model card | |
| Phare | 64.3% | #36 of 62 | Official leaderboard | |
| SEA-HELM | 64.7 | #43 of 71 | Official leaderboard | |
| StudyArena Text | 1,188 | #71 of 85 | Official leaderboard | |
| Lech Mazur Creative Writing | -2.56 | #43 of 50 | Official leaderboard | |
| Extended NYT Connections | 12.9% | #95 of 103 | Official leaderboard | |
| DesignArena ASCII Art | 1,057 | #73 of 77 | Official leaderboard | |
| EQ-Bench 4 | 983 | #32 of 32 | Official leaderboard |
Math and science
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| AIME 2025 | 86.3% | #68 of 143 | Lab model card | |
| Beyond AIME | 66.9% | #7 of 14 | Lab model card | |
| DTBench | 75.5% | #105 of 180 | Official leaderboard | |
| ProofBench | 10% | #44 of 59 | Official leaderboard | |
| Surface Evolver Bench | 26.9% | #22 of 25 | Official leaderboard | |
| ProofBench v1.1 | 9.00% | #48 of 52 | Official leaderboard |
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| SWE-bench Verified | 77.6% | #32 of 155 | Lab model card | |
| DesignArena Website | 1,135 | #106 of 151 | Official leaderboard | |
| ALE-Bench | 1,131 | #90 of 126 | Official leaderboard | |
| DesignArena Code | 1,112 | #107 of 143 | Official leaderboard | |
| DesignArena Game Dev | 1,069 | #112 of 143 | Official leaderboard | |
| DesignArena UI Component | 1,070 | #110 of 139 | Official leaderboard | |
| DesignArena 3D Design | 1,043 | #112 of 136 | Official leaderboard | |
| DesignArena Data Visualization | 1,118 | #101 of 118 | Official leaderboard | |
| Code Arena | 1,263 | #113 of 128 | Official leaderboard | |
| WebDev Arena | 1,265 | #93 of 105 | arena.ai, via Epoch AI | |
| Code Migration | 5.13% | #71 of 78 | Official leaderboard | |
| Vibe Code Bench | 2.89% | #114 of 120 | Official leaderboard | |
| DesignArena SVG | 940 | #102 of 103 | Official leaderboard | |
| DesignArena Agentic Overall | 887 | #74 of 74 | Official leaderboard |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| PACT | 1,538 | #11 of 26 | Official leaderboard | |
| tau3-bench | 63.2% | #10 of 14 | Lab model card | |
| BrowseComp | 48.6% | #65 of 81 | Lab model card | |
| Agent Arena | -13.73% | #51 of 56 | Official leaderboard |
Professional work
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LEXam-hard | 31.9% | #9 of 15 | Official leaderboard | |
| CorpFin v2 | 58.8% | #78 of 119 | Official leaderboard | |
| Vals Finance Agent | 46.1% | #34 of 49 | Official leaderboard | |
| TaxEval v2 | 68% | #95 of 129 | Official leaderboard | |
| Harvey LAB (Legal Agent Benchmark) | 0.42% | #61 of 80 | Official leaderboard | |
| MedCode | 33.8% | #78 of 96 | Official leaderboard | |
| Vals Finance Agent v2 | 32.1% | #70 of 80 | Official leaderboard | |
| MedScribe | 67.7% | #88 of 97 | Official leaderboard | |
| Legal Research Bench | 9.13% | #86 of 93 | Official leaderboard | |
| Vals Index v2 | 17.9% | #61 of 65 | Official leaderboard | |
| Tax Agent Bench | 4.48% | #81 of 84 | Official leaderboard | |
| Excel Modeling Benchmark | 11.2% | #74 of 76 | Official leaderboard | |
| MortgageTax | 28.9% | #85 of 87 | Official leaderboard | |
| CaseLaw v2 | 44.2% | #52 of 53 | Official leaderboard |
Vision, audio and robotics
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Vision Arena | 1,199 | #80 of 136 | Official leaderboard | |
| GeoBench | 56% | #38 of 48 | Official leaderboard | |
| 4DCodeBench | 0.32 | #15 of 16 | Official leaderboard |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| nocot-bench (NCRI) | 67.5 | #109 of 185 | Official leaderboard | |
| Molecular Biology Capabilities Test | 45.6% | #59 of 90 | Official leaderboard | |
| World-Class Bio | 33% | #58 of 74 | Official leaderboard | |
| AILuminate Safety 1.2 | 16.0 | #4 of 5 | Official leaderboard | |
| MLCommons AILuminate | 16% | #4 of 5 | Official leaderboard | |
| Human Pathogen Capabilities Test | 33.8% | #77 of 86 | Official leaderboard |
Other
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LMCA | 26.1% | #100 of 147 | Official leaderboard | |
| Conceptual Reasoning Index | 36.6 | #106 of 151 | Official leaderboard | |
| SAGE (Vals) | 37.6% | #62 of 81 | Official leaderboard | |
| ACCoRD | 45.6 | #119 of 152 | Official leaderboard | |
| SAGE v1.1 (Vals) | 72% | #28 of 29 | Official leaderboard | |
| Vals Multimodal Index | 34.8% | #33 of 33 | Official leaderboard |
Its lab's numbers put it at 143.5, 1.9 points above its ECI
The Self-Reported Index fits one score to the 7 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 140.0 to 147.0. How we compute it.
- COLLIE 95.8
- AIME 2025 86.3
- SWE-bench Verified 77.6
- IFBench 69
- Beyond AIME 66.9
- tau3-bench 63.2
- BrowseComp 48.6
Mistral AI's models over time
Mistral AI's models in color, against every other model on the Epoch Capabilities Index, by release date.
Source: Epoch Capabilities Index, Epoch AI (CC BY 4.0)
Its launch post
The benchmarks the lab chose to headline, with the scores it gave.