Devstral is a coding model trained to resolve repository issues through an AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition → environment that connects it to code and tests. That environment matters: the model must inspect existing software and make coordinated changes to address an issue. In its May 2025 research preview, Mistral AI reported a 46.8% score on SWE-benchThe standard benchmark for AI coding agents: real GitHub issues from real repositories, scored by whether the agent's patch passes the project's own tests.Full definition →, a collection of 500 screened software issues. The result measures issue-solving within an agent setup, providing evidence about a broader task than completing a function.
Terms in this piece · Glossary
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
SWE-bench — The standard benchmark for AI coding agents: real GitHub issues from real repositories, scored by whether the agent's patch passes the project's own tests.
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
