Mistral AI has launched Mistral Code, a new AI-powered coding assistant built specifically for enterprises, offering deployment flexibility and full-stack control. The product is open for private beta for JetBrains IDEs and VSCode, with general availability expected soon.
Mistral Code bundles intelligent code assistance, local and cloud deployment options, and enterprise-grade tools into a single package. Unlike typical SaaS-based copilots, the system is designed to run entirely within an enterprise’s secure infrastructure, on cloud, reserved capacity, or utilising on-premises GPUs.
“Every line of code resides inside the customer’s enterprise boundary,” the company said in its announcement.
At its core, Mistral Code integrates four models: Codestral for autocomplete, Codestral Embed for code retrieval, Devstral for agentic development tasks, and Mistral Medium for chat-based help. Teams can fine-tune these models or distil them into lightweight variants, a feature Mistral claims is unmatched by competitors tied to closed APIs.
Sophia Yang, head of developer relations at Mistral AI, highlighted the platform’s flexibility and deep integration on an X post, calling it “the most customisable AI-powered coding assistant for enterprises”. She pointed to its ability to automate code generation, debugging, documentation, and even migration tasks, without compromising visibility and compliance.
Early adopters include Abanca, a popular bank in Spain and Portugal, SNCF, France’s national railway company and Capgemini, each deploying Mistral Code across thousands of developers under hybrid or on-prem setups. With built-in observability, role-based access, and 24/7 support under one SLA, Mistral positions its platform as a single-vendor alternative to fragmented AI dev tools that stall at proof-of-concept.
The product builds on the open source Continue project, but adds features like audit logging, seat management, and agentic workflows that allow AI to handle complete software tickets, not just suggest lines of code.