Mistral Code is powered by four in-house AI models
Enterprises can fine-tune and post-train these AI models
Mistral Code is proficient in more than 80 programming languages
Companies such as Abanca, SNCF, and Capgemini have started using Mistral Code
Photo Credit: Unsplash/Solen Feyissa
Mistral Code, an artificial intelligence (AI) coding assistant, was introduced by the company on Wednesday. The Paris-based AI firm highlighted that the platform was built for enterprise developers, and is designed to help them increase their productivity while writing and deploying code. The platform is powered by the company's in-house AI models, and offers enterprises flexibility to adapt them to their needs. The coding assistant arrives just two weeks after Mistral released Devstral, its open-source coding agent that can perform software development-related tasks.
Mistral Code Is Powered By Four AI Models
In a newsroom post, the AI firm detailed its new coding assistant. Mistral Code is built on the open-source project Continue, and the company claims that it comes with the granular controls and transparency that enterprises require. It is currently available as private beta on JetBrains IDEs and VSCode, and it will soon be made generally available.
Mistral Code comes with several features such as an integrated development environment (IDE) assistant that exists within the virtual system, options for local deployment, and support for enterprise tools. The platform is powered by four in-house large language models (LLMs).
Codestral will enable the assistant in code completion while Codestral Embed will allow it to search and retrieve code. Similarly, Mistral Code will use Devstral for agent-based coding, and Mistral Medium for all kinds of chat assistance. Additionally, enterprises will be allowed to fine-tune or post-train the AI models on private repositories, or to distill lightweight variants depending on their requirements.
Notably, the coding assistant is well versed in more than 80 programming languages. It can also use reasoning for files, Git diffs, terminal output, and issues. Mistral said it is currently testing the platform for additional capabilities such as writing new modules, updating tests, and executing shell commands. These capabilities are said to come with configurable approvals to let developers stay in charge.
Mistral highlighted that several enterprises have already adopted and deployed Mistral Code within their organisations. These include Spain-based bank Abanca, French national state-owned railway company Société Nationale des Chemins de Fer Français (SNCF), and the French multinational IT giant Capgemini.
Mistral Releases Magistral, Its First-Ever Reasoning AI Models With Transparent Chain-of-Thought11 June 2025
Search
