devstral-latest@eu

This enterprise-oriented text model is designed for software engineering agents that must inspect codebases, work across several files, and use tools during a task. Its combination of tool calling and structured output makes it well suited to controlled agent workflows that need machine-readable results, such as code analysis, repository exploration, and multi-file editing.

The model supports a 256,000-token context window, providing substantial room for repository context and iterative agent work. The supplied evidence describes a text-in/text-out model and shows tool use plus structured output, but it does not establish a specific architecture, parameter count, training method, or benchmark performance. It is therefore best evaluated for practical software engineering tasks rather than claims about model scale or research leadership.