Mistral Medium 3.5

Mistral Medium 3.5 is Mistral AI's first flagship merged model: a dense 128B-parameter multimodal model with a 256k context window that unifies instruction-following, reasoning, and coding capabilities in a single set of weights. It replaces Mistral Medium 3.1 and Magistral in Le Chat, and Devstral 2 in the Vibe coding agent. Reasoning effort is configurable per request, and the vision encoder was trained from scratch to handle variable image sizes and aspect ratios. Released as open weights under a Modified MIT License.

Benchmark results

Benchmark Score Tags Source
AIME 2025 86.3% self-reported llm-stats link →
Beyond AIME 66.9% self-reported llm-stats link →
BrowseComp 48.6% self-reported llm-stats link →
COLLIE 95.8% self-reported llm-stats link →
IFBench 69.0% self-reported llm-stats link →
SWE-Bench Verified 77.6% self-reported llm-stats link →
Tau3 Airline 72.0% self-reported llm-stats link →
Tau3 Banking 13.4% self-reported llm-stats link →
Tau3 Retail 76.1% self-reported llm-stats link →
Tau3 Telecom 91.4% self-reported llm-stats link →