French AI startup Mistral has released a series of updates to its product portfolio to remain competitive within the highly competitive AI space.
Mistral's chatbot platform Le Chat can now search the online – with quotes in line, much like OpenAI's ChatGPT. It also includes a “canvas” tool modeled on ChatGPT Canvas that enables users to switch, transform or edit content resembling website mockups and data visualizations, leveraging Mistral’s AI models.
“You can use (the Canvas feature) to create documents, presentations, code, models… the list goes on,” Mistral writes in a single Blog Post. “You can modify content in-place without regenerating responses, versioning your designs, and previewing your designs.”
In addition, Le Chat can now process large PDF documents and pictures for evaluation and summarization, including files containing graphs and equations. Starting today, the platform integrates Black Forest Labs' Flux Pro model for image generation. And Le Chat can now host shareable automated workflows for tasks like expense report scanning and invoice processing; Mistal calls these AI “agents.”
Some of Le Chat's latest features, all of which remain free through the beta period, are made possible by Mistral's latest models.
One, Pixtral Large, can handle each text and pictures – it's the second in Mistral's Pixtral family of models. With a weight of 124 billion parameters, Pixtral Large matches or outperforms leading models on certain multimodal benchmarks, including Anthropic's Claude 3.5 Sonnet, Google's Gemini 1.5 Pro, and OpenAI's GPT-4o. (Parameters roughly correspond to a model's problem-solving capabilities, and models with more parameters generally perform higher than those with fewer parameters.)
“In particular, Pixtral Large is able to understanding documents, diagrams and natural images,” Mistral wrote in a single second blog post. “The model demonstrates image understanding at a borderline level.”
Mistral today also introduced a new edition of Mistral Large, its flagship line of text-only models. The latest model, called Mistral Large 24.11, brings “remarkable improvements” in understanding long contexts, in accordance with Mistral, making it well suited to use cases resembling document evaluation and task automation.
Both Pixtral Large and Mistral Large 24.11 will be used outside of Le Chat under two licenses: a more restrictive license for research and a company license for development and commercialization. Mistral Large 24.11 is already included in Mistral's API and on the Hugging Face AI platform and can soon be available via cloud platforms resembling Google Cloud and Microsoft Azure, Mistral says.
Paris-based Mistral, which recently raised $640 million in enterprise capital, continues to step by step expand its AI offering. In recent months, the corporate has launched a free service for developers to check its models, an SDK for patrons to fine-tune these models, and more Modelsincluding a generative model for code called Codestral.
Co-founded by Meta and DeepMind alumni, Mistral's goal is to develop highly competitive models and services around these models – and ideally generate profits while doing it. While the “earning money” part proves difficult (as is the case with most generative AI startups), Mistral allegedly began generating revenue this summer.
“At Mistral, our approach to AI is different – ​​we don’t pursue artificial general intelligence at any cost; Our mission is to as an alternative put breakthrough AI in your hands so you possibly can resolve what you wish to do with advanced AI capabilities,” the corporate wrote in considered one of its blogs today. “This approach has allowed us to be fairly economical with our capital while offering breakthrough capabilities at reasonably priced prices.”
