Devstral Medium

Last updated: Recently

Model Type

Code

Max Context

131K

Max Output

Parameter Scale

Devstral Medium is high performance code generation and agentic reasoning model developed jointly by Mistral AI and All Hands AI. 131,072 token context window. Includes independent benchmarks from Artificial Analysis.

Input Modality

text, file

Output Modality

text

Inference Speed

76.973 tokens/s

Latest Release Date

7/10/2025

SDK Ecosystem

artificial-analysismanufactureraa-bootstrap

Model Overview Fields

name

Devstral Medium

Release date

7/10/2025

Performance

76.973 tokens/s

model_creator.name

Mistral

AA Evaluation Scores

AA Intelligence Index

18.7

AA Coding Index

15.9

AA Math Index

4.7

MMLU Pro

0.708

GPQA

0.492

HLE

0.038

LiveCodeBench

0.337

SciCode

0.294

Math-500

0.707333333333333

AIME

0.0666666666666667

AIME 2025

0.0466666666666667

IFBench

0.299319727891156

LCR

0.286666666666667

TerminalBench Hard

0.0909090909090909

TAU2

0.198830409356725

Output Speed (tokens/s)

76.973 tok/s

TTFT (s)

0.519s

First Answer Token (s)

0.519s

Mistral

Devstral Medium

$0.40/M$2.00/M

There are no reviews for this model yet.

artificial-analysismanufactureraa-bootstrap

0.0