Mistral: Devstral Medium

Mistral: Devstral Medium

mistralai · Released Jul 10, 2025 Professional
Intelligence #14 / 590
82.0 Our Score
Speed #268 / 279
29.3 tokens / sec
Input #363 / 592
$0.400 per 1M tokens
Output #393 / 592
$2.00 per 1M tokens
Context #245 / 592
131,072 tokens

Analysis Summary

Devstral Medium is Mistral's mid-tier developer model, with an intelligence index of 12.4, livecodebench at 0.337, and GPQA at 0.492. Tool use and function calling are supported, and the file input modality adds document handling. The agentic index of 14.5 is modest, and the math index of 4.7 is weak, limiting its utility for quantitative tasks.

For businesses, this model fits coding assistance, code review, and lightweight agentic workflows where a mid-range model is sufficient. The 131K context window handles most codebase sizes. It is not suited to complex reasoning, mathematical analysis, or high-stakes client-facing content.

At $0.40 input and $2.00 output, it sits in a competitive mid-market tier. Teams that need reliable tool use and coding support without frontier pricing will find it a practical option, though Devstral Small 1.1 offers better value for simpler tasks at a much lower price.

Assessed June 30, 2026

Editorial notes

Mistral Devstral Medium is a coding-focused model with tool use and function calling, moderate reasoning, and competitive pricing at $0.40 input for its capability tier.

Rankings consider pricing, capabilities, benchmarks, and real-world applicability and are refreshed as new models launch. Feedback?

Performance Profile

Intelligence2.5Technical2.2Value7.3Content3.3
Intelligence 2.5/10
Technical 2.2/10
Content 3.3/10
Value 7.3/10

How Mistral: Devstral Medium compares

Mistral: Devstral Medium ranks #202 of 385 AI models we track for overall intelligence, #229 of 293 for agentic tasks. Its 131K-token context window is larger than 59% of the models we list. At $0.40 per million input tokens it is cheaper than 39% of comparable models.

About Mistral: Devstral Medium

Devstral Medium is a high-performance code generation and agentic reasoning model developed jointly by Mistral AI and All Hands AI. Positioned as a step up from Devstral Small, it achieves..

Capabilities

Tool Use Function Calling

Performance Indices

Source: Artificial Analysis

12.4 Intelligence Index
14.5 Agentic Index
4.7 Math Index

Benchmark Scores

Intelligence

GPQA Diamond 49.2% Graduate-level scientific reasoning
HLE 3.8% Humanity's Last Exam
MMLU Pro 70.8% Multi-task language understanding
MATH 500 70.7% Mathematical problem-solving
AIME 6.7% Competition mathematics
AIME 2025 4.7% Competition mathematics (2025)
SciCode 29.4% Scientific computing

Technical

LiveCodeBench 33.7% Live coding evaluation
TerminalBench Hard 9.1% Agentic terminal tasks
τ²-Bench 19.9% Conversational agent benchmark

Content

IFBench 29.9% Instruction following
LCR 28.7% Long-context reasoning

Benchmark data from Artificial Analysis and Hugging Face

How does Mistral: Devstral Medium stack up?

Compare side-by-side with other professional models.

Compare Models

Model Information

OpenRouter ID mistralai/devstral-medium
Providermistralai
Release Date July 10, 2025
Context Length131,072 tokens
Status Active

Pricing

Token Type Cost per 1M tokens Cost per 1K tokens
Input $0.40 $0.000400
Output $2.00 $0.002000

Leaderboard Categories

Frequently asked questions about Mistral: Devstral Medium

How much does Mistral: Devstral Medium cost?

Mistral: Devstral Medium costs $0.40 per million input tokens and $2.00 per million output tokens.

What is the context window of Mistral: Devstral Medium?

Mistral: Devstral Medium has a context window of 131,072 tokens (131K).

What can Mistral: Devstral Medium do?

Mistral: Devstral Medium supports tool use and function calling.

Who created Mistral: Devstral Medium?

Mistral: Devstral Medium is developed by Mistral and was released on July 10, 2025.