Mistral AI has released Codestral 25.01, an upgraded coding model featuring a more efficient architecture and improved tokenizer that generates and completes code approximately two times faster than the original version.

  • The model is now the clear leader for coding in its weight class and achieves state-of-the-art results for fill-in-the-middle (FIM) use cases across benchmarks.
  • Codestral 25.01 debuts at number one on the LMsys copilot arena leaderboard.
  • It is available via IDE partners like Continue for VS Code and JetBrains, as well as through APIs on la Plateforme, Google Cloud Vertex AI, and Azure AI Foundry.

The update provides developers with more precise suggestions and significantly reduced latency, which Mistral AI and partner Ty Dunn describe as critical components for accurate and efficient software development.