French artificial intelligence startup Mistral has unveiled its first generative AI model for coding, called Codestral. The model is designed to assist developers by completing coding functions, writing tests, and filling in partial code.
It can also answer questions about a codebase in English. According to Mistral, Codestral has been trained on over 80 programming languages, including Python, Java, C++, and JavaScript.
The model has 22 billion parameters, which influence its ability to analyze and generate text. However, running Codestral effectively requires substantial computing power.
While Mistral describes the model as “open,” there are significant restrictions on its use.
Unveiling Codestral for coding assistance
The startup’s license prohibits the use of Codestral and its outputs for commercial activities.
A provision allows for “development,” but it explicitly bans “any internal usage by employees in the context of the company’s business activities.” Despite these limitations, generative AI tools like Codestral are becoming increasingly popular among developers.
A recent Stack Overflow survey found that 44% of developers currently use AI tools in their development processes, with another 26% planning to do so soon. However, these tools are not without their drawbacks.
An analysis by GitClear found that generative AI tools have contributed to errors in codebases, and security researchers have noted that they can introduce vulnerabilities in software projects. Mistral has launched a hosted version of Codestral on its Le Chat conversational AI platform, along with a paid API.
The startup has also integrated Codestral into app frameworks and development environments such as LlamaIndex, LangChain, Continue.dev, and Tabnine. As developers weigh the benefits and drawbacks of generative AI tools, the release of Codestral adds to the ongoing conversation about the role of AI in coding and software development.
