Mistral Medium 3.5 — benchmark results

Mistral's open 128B multimodal Medium 3.5 model unifying chat, reasoning, and coding. Provider: Mistral. Released 2025-12-02. Access: API.

Unified ELO 1635 ± 17, rank #403 of 1776 rated models, from 96 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
AA TAU-2 Bench94.15Accuracy (%)92
UGI Leaderboard49.47UGI Score89
UGI - Natural Intelligence40.39NatInt Score87.5
UGI - Writing45.47Writing Score86.3
AA IFBench68.78Accuracy (%)82.6
AA Terminal-Bench Hard33.33Accuracy (%)77.7
Chatbot Arena (Text)1427Elo77.1
AA Omniscience - Health27.3Accuracy (%)76.3
Artificial Analysis Intelligence Index29.95Intelligence Index75.1
AA Omniscience - Software Engineering (SWE) - Swift52Accuracy (%)74.8
AA Long Context Reasoning61Accuracy (%)73.8
AA Omniscience - Humanities & Social Sciences26.7Accuracy (%)72.1

Interactive version: theaggregate.ai/model?slug=mistral-medium-3-5 · How the rankings work · Data refreshed daily, snapshot 2026-07-22.