Devstral Medium
Mistral's mid-tier agentic coding model — 131K context, tool calling, $0.40/$2 per 1M tokens.
About
Devstral Medium is Mistral AI's proprietary agentic coding model, released July 10, 2025 as the bigger sibling of the open-weight Devstral Small. It is built for software-engineering agents: long-horizon coding tasks, function calling, structured output, and multi-step automation pipelines, with a 131K-token context window.
It is priced for volume agent work at $0.40 per million input tokens and $2.00 per million output tokens. Independent signal: Artificial Analysis Intelligence Index 9.0 overall. Mistral reports strong SWE-Bench Verified results, but the exact figure cited in coverage traces back to Mistral's own materials — treat as provider-claimed. It is served through Mistral's La Plateforme API and is proprietary, unlike the Apache-licensed Devstral Small.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Independent (Artificial Analysis Intelligence Index): 9.0 overall, math 4.7.
Provider-reported: strong SWE-Bench Verified score (exact figure from Mistral materials only).
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 7/10
Release date, 131K context, and pricing corroborated across independent API directories (Puter, Galaxy.ai) with consistent figures. Docked for: (1) headline SWE-Bench claims are provider-reported with no exact independent figure surfaced; (2) proprietary — unlike Devstral Small, no weights; (3) modest independent benchmark footprint beyond the Artificial Analysis index.