Mistral Medium 3
- Made by
- Mistral
- Released
- Announced in
- Medium is the new large.mistral.ai
- Event
- Mistral Medium 3
- API record
- /api/v1/models/mistral-medium-3
Benchmark results
| Benchmark | Score | Source check |
|---|---|---|
| GPQA Diamond5-shot; chain of thought | 57.1% | CheckedA person checked this number against its source on 20 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| shots and chain of thought not stated · reported later, in Magistral | 59.6% | CheckedA person checked this number against its source on 20 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| AIME 2024No tools · pass@1 · reported later, in Magistral | 26.8% | CheckedA person checked this number against its source on 20 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| majority vote over 4 samples · parallel test-time compute | 34.4% | CheckedA person checked this number against its source on 20 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| majority vote over 64 samples · parallel test-time compute | 43.3% | CheckedA person checked this number against its source on 20 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| AIME 2025No tools · pass@1 · reported later, in Magistral | 21.2% | CheckedA person checked this number against its source on 20 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| majority vote over 4 samples · parallel test-time compute | 25.8% | CheckedA person checked this number against its source on 20 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| majority vote over 64 samples · parallel test-time compute | 30.0% | CheckedA person checked this number against its source on 20 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| MMLU-Pro5-shot; chain of thought | 77.2% | CheckedA person checked this number against its source on 20 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| LiveCodeBenchLiveCodeBench v6; 0-shot | 30.3% | CheckedA person checked this number against its source on 20 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
| LiveCodeBench v5 · reported later, in Magistral | 29.1% | CheckedA person checked this number against its source on 20 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced. |
Sources
Reported by Mistral
- Medium is the new large.mistral.ai · 7 May 2025
Original image - Magistralmistral.ai · 10 Jun 2025
Original image
