A major UI update from Mistral AI
Mistral is finally getting serious about the user experience, and has just released a major update to their chat. Welcome, Le Chat!
I've always been a fan of Mistral AI for many reasons. Their early models, which were small and open-source, truly pushed the frontier of AI. I also admire their status as a small startup taking on the big players, and being French added a certain charm and geopolitical significance. Earlier this year, we had the pleasure of hosting one of the co-founders, CTO Timothée Lacroix, as a guest at AMLD at EPFL (you can see his talk and fireside chat here), which was incredibly insightful. However, I've always wondered why they didn't focus more on the user interface side.
Just this weekend, I was discussing with a colleague at a party about how, despite my admiration for Mistral models, I rarely use them due to their lack of easy accessibility through chat interfaces. Well, today Mistral AI delivered a massive upgrade in beta, and it's truly impressive.
New Features in Mistral AI's Chat
Web Search with Citation: The new version includes web search capabilities with proper citations, making it easier to find and reference information.
Canvas for Ideation: This feature allows users to brainstorm and visualize ideas more effectively.
Image and Document Uploads: You can now upload images and documents for the AI to understand and process. This is powered by a new state-of-the-art vision model, Pixtral Large.
Image Generation: Powered by Black Forest Labs Flux Pro, this feature allows for high-quality image generation directly from the web browser.
I took a moment to play around with the new features, and I must say, I'm impressed. The image generation, in particular, blew me away. I've seen many Black Forest Labs images before, but they were always a bit complicated to generate. Having this functionality out of the box in a web browser is truly amazing.
While I haven't interacted much with the image understanding feature, it seems reasonably good given Mistral's benchmarks. The canvas for ideation is also nice addition, though I personally don't use these functionalities much even in other platforms like Claude or ChatGPT.
Incredible Image Generation Results
I tried a few prompts for image generation and was truly blown away by the results. Below, you can see the different versions between OpenAI's image generation and Mistral's image generation. The quality difference is stark. To be seen how OpenAI will respond, but Mistral is now my go-to place for image generation.
This update is a really cool addition, and amazingly, it's currently free! It feels like Mistral AI is seriously entering the arena, and we can certainly use more competition. Personally, I would love to see some work on a mobile client with voice recognition and perhaps a desktop client, especially for work that eventually also has desktop recognition built-in.
If Mistral AI keeps up this level of quality, I'll be a happy paying user. Awesome job, Mistral AI! If you want to check it out, just go to https://chat.mistral.ai/.
CODA
This is a newsletter with two subscription types. I highly recommend to switch to the paid version. While all content will remain free, I will donate all financial support to the EPFL AI Center.
To stay in touch, here are other ways to find me:
Social: I’m mainly on LinkedIn but also have presences on Mastodon, Bluesky, and X.
Writing: I infrequently write on another Substack on digital developments in health, called Digital Epidemiology.