Mistral Medium 3.5 (High) — benchmark results
Provider: Mistral. Released 2025-12-02. Access: API.
Unified ELO 1623 ± 22, rank #425 of 1752 rated models, from 13 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| PACT (Lechmazur) | 1538 | CMS Points | 60 |
| Finance Agent v1.1 | 46.11 | Score (self-reported) | 29.1 |
| Vals AI ProofBench | 10 | Accuracy (%) | 24.7 |
| MedCode | 33.75 | Score (self-reported) | 20.8 |
| Buyout Game (Lechmazur) | 1332.2 | Bradley-Terry Rating | 20 |
| NYT Connections Extended | 27.8 | Score (%) | 17.4 |
| Finance Agent v2 | 32.1 | Score (self-reported) | 16.7 |
| Multi-turn Debate (Lechmazur) | 1398.1 | Bradley-Terry Rating | 14.7 |
| MedScribe | 67.73 | Score (self-reported) | 9.6 |
| Vals Index | 37.81 | Score (self-reported) | 8.3 |
| Vals Multimodal Index | 34.77 | Score (self-reported) | 5.3 |
| Sycophancy (Lechmazur) | 22.1 | Sycophancy rate % (lower is better) | 3.2 |
Interactive version: aibenchmarks.dev/model?slug=mistral-medium-3-5-high · How the rankings work · Data refreshed daily, snapshot 2026-07-16.