Content
summary Summary

Mistral AI has launched Mistral Code, an enterprise-focused coding assistant designed to give companies more control and security than existing solutions.

Ad

Mistral Code brings together four of the company's specialized AI models: Codestral for automatic code completion, Codestral Embed for code search and retrieval, Devstral for agent-based coding tasks, and Mistral Medium for chat-based support.

The tool supports over 80 programming languages and builds on the open source Continue project. It's currently available as a private beta for JetBrains and Visual Studio Code.

The company is currently testing expanded capabilities that push beyond typical code assistance. The system aims to process complete development tickets independently - from opening files and creating modules to updating tests and executing command line operations. To maintain oversight, senior developers can configure specific release processes.

Ad
Ad
Screenshot: AI coding assistant interface shows corrected code, error list, and chat input for problem solving.
istral Code helps speed up development by assisting programmers with tasks like error detection and bug fixing. | Image: Mistral

Addressing enterprise challenges

Mistral AI says it created Mistral Code to address four main pain points enterprise customers have with existing coding assistants: poor integration with proprietary code repositories, limited ways to customize underlying models, shallow feature sets, and fragmented responsibilities split between multiple vendors.

Companies can choose from three deployment options: through the cloud, on dedicated cloud servers, or on their own isolated hardware for sensitive applications. The system allows companies to customize its models using their internal data. An administrative interface provides control over user permissions, tracks system usage, and maintains compliance records.

Early adopters include Abanca bank, which runs a split system - testing new code in the cloud while keeping critical banking software on internal servers. France's SNCF railway has deployed the tool to its 4,000-person development team, while Capgemini plans to roll it out to 1,500 developers working in regulated sectors.

Coding remains one of the most promising uses for large language models, even though issues like hallucinations persist. Investors are pouring billions into specialized AI coding tools such as Cursor, and OpenAI is reportedly looking to acquire its competitor Windsurf for a similar amount.

Mistral's expansion into enterprise-grade coding tools signals growing competition in the AI-assisted development space. This market is rapidly evolving into a fierce battleground, mirroring the intense rivalry seen in the chat tool sector among model providers. Companies are investing billions to develop these models, only to see them quickly become commoditized.

Recommendation

At best, they're operating on razor-thin margins, but more likely, they're hemorrhaging money as cutthroat competition drives service prices to unsustainable lows. For example, Mistral's new Medium 3 model, aimed at enterprise customers, claims over 90% of the performance of Anthropic's Claude Sonnet 3.7 at a much lower price.

Ad
Ad
Join our community
Join the DECODER community on Discord, Reddit or Twitter - we can't wait to meet you.
Support our independent, free-access reporting. Any contribution helps and secures our future. Support now:
Bank transfer
Summary
  • Mistral AI has introduced Mistral Code, a coding assistant designed for businesses that emphasizes enhanced control and safety compared to existing tools.
  • The assistant uses four purpose-built AI models, supports over 80 programming languages, and is launching as a private beta for JetBrains and Visual Studio Code.
  • Mistral Code can run in the public cloud, as a dedicated cloud setup, or fully isolated on a company's own servers, and allows training using internal company data.
Sources
Jonathan writes for THE DECODER about how AI tools can make our work and creative lives better.
Join our community
Join the DECODER community on Discord, Reddit or Twitter - we can't wait to meet you.