Mistral Medium 3: benchmark results
Mistral's enterprise multimodal mid-tier positioned near Claude 3.7 Sonnet performance at much lower cost, deployable on-prem (May 2025). Provider: Mistral. Released 2025-05-07. Access: API.
Aggregate Rating 1582 ± 9, rank #500 of 1656 rated models, from 265 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| BlueBench - Legal | 67.17 | Score (%) | 100 |
| UGI Leaderboard | 56.45 | UGI Score | 97 |
| Grip on LLMs - Dutch Factuality Average | 75 | Accuracy (%; mean of three GP-IRT estimates, Dutch machine t | 96.7 |
| Grip on LLMs - Dutch tinyARC | 82 | Accuracy (%; GP-IRT estimate from 100 tinyBenchmarks items, | 93.3 |
| Grip on LLMs - Dutch tinyTruthfulQA | 76 | Accuracy (%; GP-IRT estimate from 100 tinyBenchmarks items, | 93.3 |
| BlueBench - QA Finance | 33 | Score (%) | 91.2 |
| Grip on LLMs - Dutch tinyMMLU | 66 | Accuracy (%; GP-IRT estimate from 100 tinyBenchmarks items, | 90 |
| BlueBench - Chatbot Abilities | 94.17 | Score (%) | 82.4 |
| BlueBench - Entity Extraction | 74.21 | Score (%) | 82.4 |
| BlueBench - RAG General | 52.57 | Score (%) | 82.4 |
| UGI - Natural Intelligence | 35.26 | NatInt Score | 81.6 |
| BlueBench - Reasoning | 75.5 | Score (%) | 79.4 |
Interactive version: theaggregate.ai/model?slug=mistral-medium-3 · How It Works · Data refreshed daily, snapshot 2026-10-20.