Mistral launches Codestral, a generative AI model for coding – NewsBytes

author
2 minutes, 21 seconds Read

Ads


World’s Leading High-rise Marketplace

Codestral can perform various tasks related to coding
May 30, 2024

11:18 am

What’s the story

Mistral, a French AI startup backed by Microsoft and valued at $6 billion, has unveiled its first generative AI model for coding, called Codestral.

As the name suggests, this model is designed to assist developers in writing and interacting with code.

It was trained on over 80 programing languages such as Python, Java, C++, JavaScript, and others, as explained by Mistral in a blog post.

Features & limitations

Capabilities and commercial usage restrictions

Codestral can complete coding functions, “fill in” partial code, write coding tests, and answer questions about a codebase (in English).

However, Mistral’s license prohibits using Codestral and its outputs for commercial activities. Codestral was possibly trained partly on copyrighted content, an assumption that might explain the startup’s usage restrictions of the AI model.

Codestral boasts 22 billion parameters that define its skill in tasks such as analyzing and generating text.

AI adoption

The debate over AI tools in development process

The release of Codestral is likely to fuel ongoing debates about wisdom of relying on code-generating models as programming assistants.

A Stack Overflow poll from June 2023 revealed that 44% of developers are now using AI tools in their development process while 26% plan to do so.

However, these tools have been found to lead to more erroneous code being pushed to codebases.

Security researchers also warn that such tools can amplify existing bugs and security issues in software projects.

Information

Codestral integrated into Mistral’s conversational AI platform, app frameworks

Despite concerns about incorrect coding, Mistral has launched a hosted version of Codestral on its conversational AI platform, Le Chat, and its paid API. The company has also integrated Codestral into app frameworks and development environments like LlamaIndex, LangChain, Continue.dev, and Tabnine.

This post was originally published on 3rd party site mentioned in the title of this site

Similar Posts