Mistral Enhances AI Offerings with New Web-Searching Chatbot and Advanced Image Processing Models

French AI startup Mistral has rolled out some exciting updates to its product lineup. They’re working hard to stay competitive in the fast-paced AI market.

First up, Mistral’s Le Chat chatbot platform can now search the web. It even provides inline citations, similar to what you see with OpenAI’s ChatGPT. Plus, there’s a new “canvas” tool. This feature lets users modify and edit content like webpage mock-ups and data visualizations using Mistral’s AI models.

As Mistral explains, “You can use [the canvas feature] to create documents, presentations, code, mockups… the list goes on. You can change the content right there without needing to regenerate responses. You can version your drafts and preview your designs.”

But that’s not all. Le Chat can now handle large PDF documents and images for analysis and summarization. This includes files with graphs and equations. They’ve also integrated Black Forest Labs’ Flux Pro model for image generation. And now, Le Chat can automate tasks like expense reporting and invoice processing, which they call “agents.”

Many of these new features are available for free while in beta. They’re made possible by Mistral’s latest models.

One of these is Pixtral Large. It can process both text and images and is the second model in the Pixtral family. With 124 parameters, Pixtral Large competes well against top models like Anthropic’s Claude 3.5 Sonnet, Google’s Gemini 1.5 Pro, and OpenAI’s GPT-4o. More parameters usually mean better performance, and this model delivers.

Mistral noted, “Pixtral Large is able to understand documents, charts, and natural images. The model shows frontier-level image understanding.”

They’ve also launched a new version of their flagship text-only model, Mistral Large, called Mistral Large 24.11. This new model brings significant improvements in understanding long contexts, making it ideal for document analysis and task automation.

Both Pixtral Large and Mistral Large 24.11 come under two licenses. There’s a more restrictive license for research and academic use, and an enterprise license for experimentation, testing, and commercialization. Mistral Large 24.11 will soon be available on cloud platforms like Google Cloud and Microsoft Azure.

Based in Paris, Mistral recently raised $640 million in venture capital. They’re gradually expanding their AI offerings. In recent months, they launched a free service for developers to test their models, an SDK for fine-tuning those models, and a new generative model for code called Codestral.

Co-founded by former employees from Meta and DeepMind, Mistral aims to create highly competitive models and services. They’re also focused on making money, which has been a challenge for many generative AI startups. However, Mistral reportedly started generating revenue this summer.

In their blog, Mistral shared, “At Mistral, our approach to AI is different. We’re not chasing artificial general intelligence at all costs. Our mission is to place frontier AI in your hands, so you can decide what to do with advanced AI capabilities. This approach has helped us be frugal with our capital while consistently delivering cutting-edge capabilities at affordable prices.”