Unlocking Coding Potential: Mistral Launches Game-Changing Devstral AI Model | TechCrunch

AI startup Mistral recently unveiled a new coding model called Devstral. Developed with All Hands AI, this model is freely available under an Apache 2.0 license, which allows commercial use without restrictions. Mistral claims that Devstral outshines popular models like Google’s Gemma 3 and DeepSeek’s V3 in coding skill benchmarks.

Devstral is noted for its ability to navigate codebases and edit multiple files, enhancing the work of software engineering agents. It can run on a single Nvidia RTX 4090 or a Mac with 32GB RAM, making it accessible for local and on-device use.

The popularity of AI coding tools is rising. Companies like JetBrains and Google have introduced similar models recently. Still, AI tools face challenges. They often produce code with vulnerabilities and errors due to their limited grasp of programming logic. Despite these issues, a recent survey showed that 76% of developers either used or planned to use AI tools in their workflows.

Mistral’s previous release, Codestral, was a generative model for code but had restrictions on commercial applications. In contrast, Devstral is marketed as a “research preview” and can be accessed through AI platforms like Hugging Face. Its pricing is $0.1 for every million input tokens and $0.3 for output tokens—tokens are essentially pieces of data that AI models interpret.

Mistral indicates that they are working on a more advanced coding model, promising to roll it out soon. Although Devstral has 24 billion parameters—affecting its problem-solving skills—Mistral’s history is noteworthy. Founded in 2023, the company aims to deliver various AI services and has raised over €1.1 billion (about $1.24 billion) with clients like BNP Paribas and AXA.

Their recent launches also include Mistral Medium 3, a versatile general-purpose model, and Le Chat Enterprise, a chatbot service for businesses that connects with tools like Gmail and Google Drive.

AI continues to shape the tech landscape, and tools like Devstral are pushing the boundaries of how software is developed. According to Gartner, by 2025, over 70% of software development will involve AI in some capacity, illustrating the growing trend towards AI-assisted programming.

As AI coding tools evolve, user feedback on platforms like Twitter shows a mix of excitement and caution. Developers appreciate the speed AI brings but express concern about reliability and code quality. The future will likely see these tools refined to address such challenges while maintaining the productivity boost they offer.



Source link

mistral