My kind of frenship 🌈

officially written 28 pages over two weekends (only have around 3-4 pages left) about everything post-training (RLHF, SFT, DPO, MPO, GRPO, also stuff like LoRA, DoRA, QLoRA, bnb) and I did it while I was sick you can just write things

Romantic star-gazing with reachy mini -> unlocked!

After years of struggle, I can finally write mypy compliant code.

Funny how hard going from laggy but works to smooth can be. Clearly the 80-20 rule still stands.

Let’s goooo! We’ve just launched a new Computer Use Agent (CUA) powered by open models, Hugging Face smolagents and @E2B for secure computer sandboxing! We're building something different. Open. Transparent. Yours. Check this out

hahaha this is totally Reachy in his gym bro arc after a bad break up

Super cool strategy to train robots, this is thinking outside of the box, congrats!

Just dropped: 🎉 NVIDIA Nemotron-Parse v1.1 Next-gen OCR for parsing PDFs & PPTs into structured, machine-ready output (text + bounding boxes + semantic classes). Ready for commercial use and to generate datasets🚀 Check the examples on Hugging face! huggingface.co/nvidia/NVIDIA-…

PRO for PROs Nano Banana PRO is available at no cost for Hugging Face PRO subscribers on Spaces, go bananas 🍌

Late night coding, brain fried. Pasted logs + code to ChatGPT and Gemini. ChatGPT explained the issue and how to solve it. Gemini told me I must have pasted the wrong code because it had nothing to do with the logs. Gemini was right.

I'm keeping a track of real-time vision models (mostly detectors) on Hugging Face we have RT-DETR, YOLO, RF-DETR and D-FINE for now what other models should we add?

icymi Hugging Face dropped a computer use agent last week 🔥 built on various Qwen3-VL models & E2B sandbox, ask the app to do anything 🙌🏻 it exposes each thinking step to you, try different models with a neat UI🤩

For the first time, in 2025, Chinese model developers surpassed the U.S. in adoption, driven by DeepSeek and Qwen. The global race for AI isn’t settled.

Not even doodle works anymore without a premium account :(

Introducing "The Eiffel Tower Llama"!🗼 Remember Golden Gate Claude? Unfortunately Anthropic's viral demo was shut down after 24h, and key technical details remained hidden. So we recreated it, uncovering key insights on steering LLMs using SAEs⚒️ Full blog post + live demo 👇

Not to diminish anyone’s achievements, but most of the things Jesus did were basically extensions of techniques Jürgen Schmidhuber published decades earlier. Turning water into wine? They had liquid-to-liquid transformation networks in ’78 BCE. As usual: great work, zero citations.

Transformers v5's first release candidate is out 🔥 The biggest release of my life. It's been five years since the last major (v4). From 20 architectures to 400, 20k daily downloads to 3 million. The release is huge, w/ tokenization (no slow tokenizers!), modeling & processing.

Really cool stuff! Voice AI is far from solved and this team is going to raise the bar, exciting times ahead!

We heard your feedback 🙏 Killer new feature on Hugging Face = duplicate any dataset to your account. Thanks to Xet, duplicating a 1TB dataset takes 2 seconds HF is going to be the best platform for large data storage