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.

Unified ELO 1527 ± 1, rank #556 of 1392 rated models, from 75 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
SpeechMap Compliance92.6% Requests Completed94.3
CritPt30Accuracy (self-reported)80.8
HumainE Leaderboard30.88HumainE Score73.6
SnakeBench25.8TrueSkill Rating70.2
BenchTable55.9Total Score (%)63.5
AA Omniscience - Science, Engineering & Mathematics31Accuracy (%)60.7
AI for Education Pedagogy - Social studies80.91Accuracy (%)59.5
AA Omniscience-26.93Score58.9
AA Omniscience - Health22.2Accuracy (%)56.9
AA Omniscience - Business18.1Accuracy (%)56.4
AA Humanity's Last Exam9.82Accuracy (%)53.4
AA CritPt0.29Accuracy (%)53.2

Interactive version: theaggregate.ai/model?slug=magistral-medium · How It Works · Data refreshed daily, snapshot 2026-09-05.