Magistral Medium: benchmark results
Mistral's first frontier-class reasoning model, the larger API-only counterpart to Magistral Small (June 2025). Provider: Mistral. Released 2025-06-10. Access: API.
Aggregate Rating 1585 ± 15, rank #494 of 1656 rated models, from 130 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| SpeechMap Compliance | 92.6 | % Requests Completed | 94.7 |
| HUMAINE | 3.69 | Overall score | 81.8 |
| AEPC-QA (Closed-Book) | 61.81 | Accuracy (%) answering from the model's own knowledge on AEP | 74 |
| HumainE Leaderboard | 30.88 | HumainE Score | 73.6 |
| AA-Omniscience Hallucination Rate | 59.33 | Hallucination Rate (%) | 72.4 |
| SnakeBench | 25.8 | TrueSkill Rating | 70.2 |
| BenchTable | 55.9 | Total Score (%) | 63.5 |
| SuperCLUE General (July 2025) - Code Generation | 75.45 | Score | 63 |
| BenchTable - Reasoning | 49.2 | Weighted Score (%) | 62.2 |
| SuperCLUE General (July 2025) - Precise Instruction Following | 23.8 | Score | 60.9 |
| AA LiveCodeBench | 52.7 | Pass@1 (%) | 59.4 |
| AA-Omniscience Index - Business | -19.6 | Omniscience Index | 58.9 |
Interactive version: theaggregate.ai/model?slug=magistral-medium · How It Works · Data refreshed daily, snapshot 2026-10-20.