Mistral AI has launched Mistral Code, an AI coding assistant for enterprises that bundles its models, an in-IDE assistant, local deployment options and admin tooling in one package, the company said on 4 June 2025.
Mistral Code is built on the open-source Continue project and is in private beta for JetBrains IDEs and VS Code. It can be deployed in the cloud, on reserved capacity or on air-gapped on-premises GPUs, keeping code inside the customer's boundary.
It is powered by four Mistral models: Codestral for autocomplete, Codestral Embed for code search, Devstral for agentic coding and Mistral Medium for chat. Customers can fine-tune the models on private repositories. An admin console provides platform controls, observability, seat management and usage analytics.
Spanish bank Abanca has deployed Mistral Code for development teams in a hybrid setup, prototyping in the cloud while core banking code stays on-premises. French railway SNCF is rolling it out to 4,000 developers, and Capgemini plans to deploy it on-premises for more than 1,500 developers working on client projects in regulated industries.
Mistral AI said it is also testing the product on completing fully scoped tickets under configurable approval workflows.