Mistral Medium 3.5

Mistral AI🇫🇷 France
active
Context window256K tokens
Input / 1M tokens$1.5
Output / 1M tokens$7.5

Version History

3.5minor

Mistral Medium 3.5 merges instruction-following, reasoning, and coding into a 128B dense model with 256k context. Released as open weights under modified MIT license with configurable reasoning effort and new vision encoder.

Benchmark Scores

Full leaderboard →
86.3%
AIME 2025
1427.0 elo
Arena Elo
74.8%
GPQA
81.6%
Hallucination Rate
118.0 tokens_per_sec
Speed (tok/s)
77.6%
SWE-bench Verified

Coverage