Magistral Medium 1

Creator

Mistral AI

Released

2025-06-10

Intelligence

12.5

Artificial Analysis Index

Coding

Artificial Analysis Index

In $/1M

$0.00

input tokens

Out $/1M

$0.00

output tokens

Blended $/1M

$0.00

3:1 blended

Speed

0

tokens / sec

Profile

License, openness, and modality — the governance layer.

License

Apache 2.0

commercial OK

Weights

Open

downloadable

Modalities

Parameters

24B

total

Trained solely with reinforcement learning on top of Mistral Medium 3, Magistral Medium is a reasoning model that achieves strong performance on complex math and code tasks without relying on distillation from existing reasoning models. The training uses an RLVR framework with modifications to GRPO, enabling improved reasoning ability and multilingual consistency.

Capability profile

Category strength across 20 domains, via LLM Stats.

biology
70physics
70chemistry
70code
50
math
50general
50reasoning
50vision
10
Via LLM Stats

Benchmark breakdown

Independent evaluation scores, normalized to 0–100.

GPQA Diamond
68
Humanity’s Last Exam
10
MMLU-Pro
75
SciCode
30
LiveCodeBench
53
MATH-500
92
AIME 2025
40
τ²-Bench (agentic)
23
Terminal-Bench Hard
9
IFBench
25
Via Artificial Analysis
© 2026 NYSGPT2525 LLC