French AI startup Mistral has released a number of updates to its product portfolio to stay competitive in the cutthroat AI field.
Mistral's Le Chat chatbot platform, like OpenAI's ChatGPT, can now enumerate quotes and search the web. It also adds a “canvas” tool along the lines of ChatGPT Canvas, which allows users to leverage Mistral's AI models to modify, transform, and edit content such as web page mockups and data visualizations. .
“You can use it.” [the canvas feature] To create documentation, presentations, code, mockups…the list goes on,” Mistral wrote in a blog post. “You can change content on the fly, version drafts, and preview designs without regenerating responses.”
In addition to all this, Le Chat can now process large PDF documents and images for analysis and summarization, including files containing graphs and equations. The platform currently incorporates Black Forest Labs' Flux Pro model for image generation. Le Chat can also now host shareable automated workflows for tasks like scanning expense reports and processing invoices. Mr. Mistal calls these AIs “agents.”
Some of Le Chat's new features, all of which remain free during beta, are made possible by Mistral's new model.
One Pixtral Large can handle both text and images. This is the second model in Mistral's Pixtral family. Boasting 124 billion parameters, Pixtral Large matches or bests leading models such as Anthropic's Claude 3.5 Sonnet, Google's Gemini 1.5 Pro, and OpenAI's GPT-4o in certain multimodal benchmarks. (Parameters roughly correspond to the model's problem-solving skills, and models with more parameters generally perform better than models with fewer parameters.)
“Specifically, Pixtral Large can understand documents, graphs, and natural images,” Mistral wrote in a second blog post. “This model demonstrates frontier-level image understanding.”
Mistral today also announced a new version of its flagship text-only model, Mistral Large. The new model, called Mistral Large 24.11, offers “notable improvements” in understanding long contexts and is suitable for use cases such as document analysis and task automation, Mistral said.
Both Pixtral Large and Mistral Large 24.11 are available for use outside of Le Chat under two licenses. One is a more restrictive license for research, and the other is an enterprise license for development and commercialization. Mistral Large 24.11 is already integrated into Mistral's API and AI platform Hugging Face, and will soon be available through cloud platforms such as Google Cloud and Microsoft Azure, Mistral said.
Paris-based Mistral recently raised $640 million in venture capital and continues to gradually expand its AI services. Over the past few months, the company has launched new models that include free services for developers to test models, an SDK that allows customers to tweak those models, and a code generation model called Codestral.
Co-founded by Meta and DeepMind alumni, Mistral's mission is to create highly competitive models and services around those models, and ideally make money in the process. While the “making money” part has proven to be difficult (as with most generative AI startups), Mistral reportedly started turning a profit this summer.
“At Mistral, we take a different approach to AI. We are not chasing artificial general intelligence at all costs. Our mission is to put cutting-edge AI in the hands of users and provide advanced AI capabilities. We want you to decide how you want to use it,” the company wrote in one of its blogs today. “This approach has allowed us to be very frugal with capital while always offering frontier capabilities at an affordable price point.”