THE BIG ONE
Gradium Launches Real-Time Speech Translation Models — Gradium has released two impressive real-time speech translation models, stt-translate and s2s-translate, that outperform existing solutions like gpt-realtime-translate in both accuracy and latency. Supporting multiple languages (English, French, German, Spanish, and Portuguese) across 20 pairs, these models are game-changers for anyone needing fast, reliable translations during conversations or presentations. I’ve already tested these in my weekend hackathon project, and the results were stellar! Read more
QUICK HITS
Baidu Releases Unlimited OCR — This new 3B-parameter model can parse entire documents in one go, thanks to its innovative Reference Sliding Window Attention. If you deal with long documents, this could save you tons of time. Learn more
Mistral OCR 4 — Mistral has upped its game with structured outputs for document parsing, making it more efficient for RAG and enterprise search applications. Perfect for when you need citation-ready documents. Check it out
Datalab Releases lift — This 9B open-weights vision model extracts structured JSON from PDFs, streamlining data extraction like never before. If you work with PDFs daily, you’ll want to give this a spin. Find out more
Nous Research Adds /learn to Hermes Agent — This new feature allows users to create workflows without writing SKILL.md by simply using slash commands. It’s a handy shortcut for streamlining your skills system. Discover it here
ONE THING TO TRY
Experiment with Gradium’s new real-time translation models in your next project. They’re perfect for enhancing communication across languages!
SIGN-OFF
That’s it for this week! Dive in and let me know what new tools you’re excited about!