Model
Model explorer

Magistral Medium

CLOSED
Mistral AI · Magistral family · released Jun 10, 2025

Closed enterprise reasoning model, RL-trained on Mistral Medium 3; 73.6% AIME2024 pass@1 (90% with maj@64), Flash Answers via Le Chat.

ReasoningCodingVisionFunction callingTool useAgentic
1485.1
Elo · rank #124
Parameters
Undisclosed
Active params
—
Context
40K tokens
Architecture
Dense transformer, reasoning RL fine-tune of Mistral Medium 3 (architecture undisclosed)
License
Proprietary
Languages
—
API price (in/out)
$2 / $5
Modalities
text
Benchmark results
Bar shows position within the tracked field; marker = field best
Aider PolyglotCoding47.1%#27
best: Claude Opus 4.5 · 89.4%
AIMEMath73.6%#94
best: GPT-5.2 · 100.0%
GPQA DiamondReasoning70.8%#123
best: GPT-5.6 · 94.6%
Humanity's Last ExamReasoning9.0%#91
best: Claude Sonnet 5 · 57.4%
LiveCodeBenchCoding59.4%#98
best: DeepSeek-V4-Pro (Think Max) · 93.5%
MATH-500Math94.3%#37
best: GPT-5 · 99.4%
Run it locally
Closed weights — available via API only. No local deployment.
Input / M tok
$2
Output / M tok
$5
Magistral family
Elo progression across releases
Magistral MediumJun 20251485.1Magistral SmallJun 20251422.4
Resources
DOCSdocs.mistral.ai/models/model-cards/magistral-medium-1-0-25-06APIAvailable via Mistral AI API
API price $2/$5 · each benchmark row carries its own source badge (see methodology)