Mistral introduces a new multimodal AI model and competes with ChatGPT with new updates to Le Chat.

French AI startup Mistral has introduced Pixtral Large: a multimodal AI model with 124 billion parameters. In addition, the Le Chat chatbot will have new features such as a web search function, an interactive canvas and image generation.

Pixtral Large: powerful multimodal model

Pixtral Large builds on previous Mistral models, including Pixtral 12-B. The model is available on the Hugging Face platform, but the license only allows non-commercial use. For commercial use, access is possible via Mistral’s own API or via a specific license request. Therefore, Pixtral Large is not fully open source, but remains accessible for research and ethical applications.

Mistral has also expanded Le Chat with new tools aimed at productivity and automation. One of the most important innovations is the citation search function, which allows users to search for information with citations in real time. The interactive canvas allows users to edit and share documents, presentations and designs in a new collaborative window.

Thanks to advanced document and image analysis, Le Chat can analyze and summarize complex PDFs and visual data. Le Chat also gets imaging with the Flux Pro model. Finally, the chatbot is given automation tools to perform repetitive tasks.

All of these features are free in the current beta version. With Pixtral Large and the expanded Le Chat, Mistral shows its ambitions to penetrate further into the global AI market.