Image: © maurice norbert/Stock.adobe.com
Trained in languages such as Python, Java and C++, the model can help software developers save time and effort when designing advanced AI applications.
French AI company Mistral has unveiled a new model called Codestral designed specifically for coding tasks.
Mistral described the model as open-weight, which means that users can access Codestral’s pretrained parameters for research and fine-tuning, but the training code and methodology, original dataset and architecture details have not been released.
The idea is to help developers save time and effort by assisting them to complete coding functions, write tests and finish partial codes with a ‘fill-in-the-middle’ mechanism. Mistral said that the Codestral AI model is trained on more than 80 programming languages, including the most popular ones such as Python, Java, C and C++.
“Codestral is an open-weight generative AI model explicitly designed for code generation tasks,” Mistral wrote in its announcement.
“It helps developers write and interact with code through a shared instruction and completion API endpoint. As it masters code and English, it can be used to design advanced AI applications for software developers.”
Request Codestral access at https://t.co/K6g5yh8m1T. It's free during a beta period of 8 weeks! pic.twitter.com/ADzcl6rALg
— Mistral AI Labs (@MistralAILabs) May 29, 2024
Based in Paris, Mistral was started a year ago by former researchers at Google’s Deepmind and Meta and is focused on creating open-source large language models to compete with the likes of OpenAI and Google.
It hit unicorn status in December after securing €385m in a round backed by Andreessen Horowitz, Nvidia and Salesforce. At the time, the start-up was valued at around $2bn.
Mistral released its most advanced language model, Mistral Large, in February. It said the new text-generation model has “top tier” reasoning capabilities and can be used for complex multilingual reasoning tasks such as including text understanding, transformation and code generation.
Around the same time, Microsoft – which has previously invested heavily in OpenAI – struck a deal with Mistral that will see the AI start-up provide commercial language models on Microsoft’s Azure cloud computing platform.
As part of the partnership, Microsoft and Mistral AI will explore collaboration around training purpose-specific models for select customers, including the European public sector.
Find out how emerging tech trends are transforming tomorrow with our new podcast, Future Human: The Series. Listen now on Spotify, on Apple or wherever you get your podcasts.
