"Hey Hunters! 👋

Mistral OCR is here, bringing lightning-fast and highly accurate text extraction to your workflow! 🚀 Whether you're processing documents, images, or real-world text, this model sets a new standard.

Who’s excited to try it? Drop your thoughts below! 🔥👇"

Mistral is cooking 🍳!

Could this also convert web page UI screenshots into AI-ready formats for agents?

For instance, could it handle dynamic layouts or interactive components such as buttons and menus to produce structured outputs that help identify which element to click?

Can't wait to try! Using OpenText OCR in the corporate world and it's a miss. Mistral continues to deliver and it looks wonderful. Congrats on the release!

really cool. will give this a try. good luck with the launch

Lightning-fast, ultra-accurate OCR?

Mistral OCR looks like it's hitting a spot for anyone dealing with scanned docs, images, or text extraction at scale...!!

If this truly delivers state of the art performance in a lightweight package, it’s a recommendation I'm going to be passing this on to everyone!!

Best of luck to the team :)

Top tier OCR product by Mistral AI whose output this year has been stellar. They released Le Chat, Codestral 25.01 and Mistral Small 3 prior to this.

A true European contender in the AI space

This is super interesting. I'll give this a try.

This looks very interesting! Looking forward to trying this soon!

I saw the performance of Mistral OCR, Gemini, Google Doc AI, and Azure OCR on your landing page. But how do their prices compare? Which one is more cost-effective for small and medium businesses?

Mistral always keeps shipping something cool

This is amazing. I have, I kid you not, been working on my own homegrown app to solve this ridiculous problem of OCR document "transforming" that leaves you with a live text but still wonky, sideways, ugly looking document. Based on the examples I'm seeing, I feel slightly better at how getting it to recognize columns and layouts wasn't just my problem, haha.

Couple clarifying questions and forgive me... I have looked but not deeply at your documentation so this may be discussed there:
1. The price is 1000 per dollar...? All your prices have a number next to the $ sign except this one. (So $0.50 for batch processing).
2. It says available via api, cloud coming soon, but then says only selective self-hosting is allowed... so could I use it via api and not self host as lay person at this stage? Or is that still to come? Forgive my ignorance on how Mistral is set up. I've mostly dealt with Anthropic .
3. Is this the correct documentation? OCR and Document Understanding | Mistral AI Large Language Models If so, It looks like a request is returned with markdown? Is there a way to change what it sends back or is markdown the only output option? And how do images get returned accurately per your examples?

Thanks again. Hats off to the crew for this!

Brett

Congratulations on launching Mistral OCR! Your text extraction solution sounds impressive with its speed and accuracy claims. Really looking forward to seeing how it performs on challenging document processing tasks. The ability to handle both digital documents and real-world text could be transformative for many workflows. Curious to hear about specific use cases where it's outperforming existing solutions. Will definitely be checking this out!

hi everyone

Good job, Mistral team - congrats! That’s a really useful product!

Whether extracting text from images or digitizing documents

Would help to know if this is the best at both. Documents can be classified into scans of offline documents, and digitally created documents. Is it the best at both?

This is such a powerful tool for anyone working with text extraction!