Magistral Medium 1.2

Released

Magistral Medium 1.2 is an AI model released by Mistral on Wednesday, Sep 17 2025, the same day as Magistral Small 1.2. It has a 128k token context windowHow much information the model can work with in one request, measured in tokens — pieces of words or other data. This includes your question, attachments, conversation and reply. Limits can vary by model variant and provider.. Benchmark results (shown below) cover LiveCodeBench, AIME 2024, AIME 2025, and GPQA Diamond.

Benchmarks

Coding

LiveCodeBenchCompetitive coding — Coding problems published so recently the AI can't have seen them in training — a contamination-free test of raw programming skill. Higher is better.
75%
#17 of 60

Reasoning & science

AIME 2024Reasoning — single-answer accuracy on the 2024 invitational mathematics exam without tools
91.82%
#5 of 21
AIME 2025Reasoning — single-answer accuracy on the 2025 invitational mathematics exam without tools
83.48%
#20 of 29
GPQA DiamondScience — Graduate-level science questions in biology, physics, and chemistry — hard enough that subject-matter PhDs score around 65%. Higher is better.
76.26%
#71 of 125

Compare Magistral Medium 1.2 with

Magistral Medium 1.2

About

Magistral Medium 1.2 appeared in Mistral's September 17, 2025 release changelog as the hosted multimodal update to Magistral Medium. It combined image understanding with the line's reasoning behavior, allowing applications to pose questions that required interpreting visual inputs and working through an answer.

The model documented a 128K-token context and was offered commercially rather than as downloadable weights. Mistral published separate results for this version on AIME, GPQA Diamond and LiveCodeBench v5. Those results distinguish the September model from the June and July Medium checkpoints, whose names and reasoning interfaces were closely related.

Frequently asked questions

Magistral Medium 1.2 was released by Mistral on Wednesday, Sep 17 2025.

All Mistral releases

45 tracked