Mistral Unveils Pixtral 12B: Europe's Answer to OpenAI's Multimodal Models?

In a move that underscores its ambition in the generative AI space, French AI startup Mistral has unveiled Pixtral 12B, its first foray into multimodal models capable of handling both text and images. This marks a significant step for the young company, which has quickly gained recognition as a potential European rival to OpenAI.

Pixtral 12B, weighing in at a substantial 24GB with 12 billion parameters, is built upon Mistral’s existing text model, Nemo 12B. The new model's capabilities extend beyond mere text comprehension, allowing it to analyze and respond to queries about images provided through URLs or base64 encoding. It's anticipated that Pixtral 12B will be able to perform tasks similar to other multimodal models, such as generating image captions or counting objects within a photograph.

Mistral has adopted an open-source approach with Pixtral 12B, making it available for download, fine-tuning, and use under the permissive Apache 2.0 license. This move aligns with the company's strategy of releasing free, open models while generating revenue through managed versions and consulting services. While web demos weren't available at the time of the model's release, Mistral has indicated plans to make Pixtral 12B accessible for testing on its chatbot and API platforms in the near future.

The release of Pixtral 12B comes on the heels of Mistral securing a substantial $645 million funding round, which valued the company at an impressive $6 billion. This injection of capital, coupled with its minority ownership by Microsoft, positions Mistral as a formidable player in the European AI landscape.

However, the development of Pixtral 12B raises questions about the image data used in its training. Like many generative AI models, Mistral's models are trained on vast amounts of publicly available data scraped from the web, often raising copyright concerns. While some companies argue that fair use rights allow for such data scraping, copyright holders have challenged this practice, leading to lawsuits against major AI firms. The specific image data used to train Pixtral 12B remains undisclosed, leaving this question open for further scrutiny.

Overall, the release of Pixtral 12B represents a significant milestone for Mistral, showcasing its technical prowess and ambition in the generative AI space. The model's open-source nature and multimodal capabilities position it as a potentially valuable tool for researchers and developers. As Mistral continues to evolve and expand its offerings, it will be fascinating to observe its impact on the AI landscape, particularly in Europe, where it's poised to challenge established players like OpenAI.

Ready to Harness the Power of AI for Your Business?

Mistral’s AI innovation is a testament to the transformative potential of artificial intelligence. At Deqode , we specialize in helping businesses leverage cutting-edge AI technologies to drive innovation and growth. Whether you're looking to develop custom AI solutions, integrate AI into your existing infrastructure, or simply explore the possibilities, our team of expert developers is ready to assist you.?

Contact us today to discover how we can help you unlock the full potential of AI for your business.

For more insights and updates on the ever-evolving world of technology, be sure to subscribe to our newsletter and follow us on X.