
Devstral Medium 2507
Mistral's proprietary mid-tier agentic coding model, 61.6% on SWE-Bench Verified.
About
Devstral Medium 2507 is the larger sibling in Mistral AI's Devstral 2507 release, a family of code-focused language models built for agentic software engineering. Where Devstral Small 1.1 shipped open-source under Apache 2.0 for local deployment, Devstral Medium is the higher-performing proprietary tier, scoring 61.6% on SWE-Bench Verified according to Mistral's announcement. It is positioned as the cost-effective step up for businesses that need stronger repository-level reasoning than the small model offers.
The model is available through Mistral's public API as devstral-medium-2507 at $0.4 per million input tokens and $2 per million output tokens, the same price point as Mistral Medium 3. It supports Mistral function calling and XML tool formats for agentic scaffolds, and enterprises can deploy it on private infrastructure or customize it through Mistral's finetuning API. It also ships inside Mistral Code for enterprise customers. Unlike the small variant, Devstral Medium has no open weights, so local and offline use is not an option.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 120; Trending score 82/100 from researcher assessment.