Mistral AI·released 2025-05-071 source
Mistral Medium 3 benchmark scores: 7 benchmarks tracked. 57% of its scores are independently reproduced — the rest are self-reported or unverified.
Leads on 0 of 7 benchmarks·best result 81.63 on MATH level 5 (reproduced)·4 of 7 independently reproduced·$0.4/$2 per M tokens
Consensus: LiteLLM · Cross-check: OpenRouter · See every model’s pricing →
| Benchmark | Score | Evidence status | Measured |
|---|---|---|---|
| MATH level 5 | 81.63accuracy (%) | reproduced· optimizedT1 | 2025-05-07Epoch AI |
| Lech Mazur Writing | 77.3accuracy (%) | unverifiedT1 | 2025-05-07lechmazur/writing Github repository |
| GPQA diamond | 46.04accuracy (%) | reproduced· optimizedT1 | 2025-05-07Epoch AI |
| DTBench | 37.12accuracy (%) | unverifiedT1 | 2025-05-07https://conceptualreasoning.ai/dtbench |
| OTIS Mock AIME 2024-2025 | 32.15accuracy (%) | reproduced· optimizedT1 | 2025-05-07Epoch AI |
| FrontierMath-2025-02-28-Private | 0.61accuracy (%) | reproduced· optimizedT1 | 2025-05-07Epoch AI |
| HLE | 0accuracy (%) | unverifiedT2 | 2025-05-07 |
Claims drawn from cited facts, not live model generation.
This model was developed by Mistral AI, placing its origin with that research lab. It was released in 2025, making it a recent arrival rather than an older-generation system.
2 cited facts
This model is scored on 7 tracked benchmarks. It holds no current top score on any of them. Its record can be trusted: the scores have been independently reproduced by outside evaluation.
3 cited facts
Across its benchmarks this model trails the top score by an average of 44.74 points, a wide gap that reads as well behind the current leaders rather than at or near the front of the pack. It holds the top score on none of the benchmarks it is measured against (sota leader count of 0), so there is no category leadership here yet.
2 cited facts
5 facts cross-checked across data sources: 2 corroborated, 1 single-source, 2 disputed.
Source facts, citations, and refresh stamp for this record.
Sources: epoch_benchmarks, litellm_prices, openrouter_models
Mistral Medium 3 is an AI model developed by Mistral AI, released 2025-05-07. Its benchmark record below tags every score as independently reproduced, self-reported, or unverified.
Mistral Medium 3 has recorded scores on 7 benchmarks, each shown with its evidence status.
Mistral Medium 3 has recorded scores on 7 benchmarks — Lech Mazur Writing, GPQA diamond, MATH level 5, OTIS Mock AIME 2024-2025, DTBench, FrontierMath-2025-02-28-Private, and 1 more. The full table above shows each score with its evidence status.
4 of 7 recorded scores (57%) are independently reproduced rather than self-reported by the lab.
Mistral Medium 3 has 7 tracked claims: 4 independently reproduced, 3 unverified.
Listed API pricing: $0.4 per million input tokens, $2 per million output tokens (prices disputed across sources). See the pricing block for the full breakdown.