We made a blog post on how you can use kernel-builder to develop and build compute kernels for the @hf.co Kernel Hub: huggingface.co/blog/kernel-...

I'd like to grow my network more on this site as it is small compared to on that other site. My interests are quite broad so drop some names to follow 👇 around these topics: #machinetranslation #MT, Universal Dependencies, #LLMs in all of its aspects (technical, ethical, data), #Dutch NLP, #OSS!

Yesterday we released support for GPT OSS (the new OpenAI open weight model) across the @hf.co ecosystem. The latest Transformers now integrates support for the kernels package and uses kernels from the HF Kernel Hub to run models like GPT OSS as fast as possible. 🚀 huggingface.co/blog/welcome...

Transformers 4.54.0 is out! This release adds support for compute kernels hosted on the Hub. When enabled, transformers can replace PyTorch layer implementations by fast, specialized kernels from the hub. github.com/huggingface/...

Just released a new version of mktestdocs. It now also supports huggingface docstrings! github.com/koaning/mkt...

Some of the ModernBERT team is back with new encoder models: Ettin, ranging from tiny to small: 17M, 32M, 68M, 150M, 400M & 1B parameters. They also trained decoder models & checked if decoders could classify & if encoders could generate. Details in 🧵:

So excited to finally release our first robot today: Reachy Mini A dream come true: cute and low priced, hackable yet easy to use, powered by open-source and the infinite community. Read more and order now at huggingface.co/blog/reachy-...

SUSE has released Cavil-Qwen3-4B, a fine-tuned, #opensource #LLM on #HuggingFace. Built to detect #legal text like license declarations, it empowers #devs to stay #compliant. #fast #efficiently. #openSUSE #AI #Licenses news.opensuse.org/2025/06/24/s...

Over the past few months, we have worked on the @hf.co Kernel Hub. Kernel Hub allows you to get cutting-edge compute kernels directly from the hub in a few lines of code. David Holz made a great writeup of how you can use kernels in your projects: huggingface.co/blog/hello-h...

Hi Berlin people! @hugobowne.bsky.social is in town & we're celebrating by hosting a meetup together 🎉 This one is all about building with AI & we'll also open the floor for lightning talks. If you're around, come hang out with us! 📆 June 16, 18:00 📍 Native Instruments (Kreuzberg) 🎟️ lu.ma/d53y9p2u

TGI v3.3.1 is released! This version switches to Torch 2.7 and CUDA 12.8. This should improve support for GPUs with compute capabilities 10.0 (B200) and 12.0 (RTX50x0 and NVIDIA RTX PRO Blackwell GPUs). github.com/huggingface/...

@aob.nl mooie tijdslijn van de stakingen in het onderwijsblad, alleen de staking van 18 maart bij de @rug.nl vergeten, wel een beetje jammer!

We just released text-generation-inference 3.3.0. This release adds prefill chunking for VLMs 🚀. We have also Gemma 3 faster & use less VRAM by switching to flashinfer for prefills with images. github.com/huggingface/...

At @hf.co we are also building...

The entire Xet team is so excited to bring Llama 4 to the @hf.co community. Every byte downloaded comes through our infrastructure ❤️ 🤗 ❤️ 🤗 ❤️ 🤗 Read the whole post to see more about these models.

Gemma 3 is live 🔥 You can deploy it from endpoints directly with an optimally selected hardware and configurations. Give it a try 👇

HuggingChat keycap sticker when?

We're thrilled to announce that development of the Orion Browser for Linux has officially started! Register here to receive news and early access opportunities throughout the development year: forms.kagi.com?q=orion_linu...

want to try QwQ-32B? it just landed on HuggingChat!

Six months after joining @hf.co we’re kicking off the first migrations from LFS -> Xet backed storage for a handful of repos on the Hugging Face Hub. A few months ago, I published a timeline of our work and this is a big step (of many!) to bring our storage to the Hub - more in 🧵👇

Followers gezocht. Nu we niet meer actief zijn op X (algemeen FS-account 150k followers) en Mastodon helaas niet het volume van het oude Twitter lijkt te krijgen, hoop ik dat BlueSky die plaats kan innemen. Social media is toch een goedkope manier om publiek te informeren. pls rt

Not only is DeepSeek R1 open, you can now run it on your own hardware with Text Generation Inference 3.1.0. Awesome work by @mohit-sharma.bsky.social and @narsilou.bsky.social !

Want to run Deepseek R1 ? Text-generation-inference v3.1.0 is out and supports it out of the box. Both on AMD and Nvidia !

🐳 DeepSeek is on Hugging Face 🤗 Free for inference! 1K requests for free 20K requests with PRO Code: https://buff.ly/4glAAa5 900 models more: https://buff.ly/40x1rua

Text-generation-inference v3.0.2 is out. Basically we can run transformers models (that support flash) at roughly the same speed as native TGI ones. What this means is broader model support. Today it unlocks Cohere2, Olmo, Olmo2 and Helium Congrats Cyril Vallez github.com/huggingface/...

🐐 DeepSeek is not on the @hf.co Hub to take part, they are there to take over! Amazing stuff from the DeepSeek team, ICYMI they recently released some reasoning models (DeepSeek-R1 and DeepSeek-R1-Zero), fully open-source, their performance is on par with OpenAI-o1 and it's MIT licensed!

Hello on the new sky!

The speed of uv is just insane. Just experimented with using it for CI of a project and installing a project, its dependencies (including Torch), and running some tests takes 30 seconds 🤯.

BIG release by DeepSeek AI🔥🔥🔥 DeepSeek-R1 & DeepSeek-R1-Zero: two 660B reasoning models are here, alongside 6 distilled dense models (based on Llama & Qwen) for the community! huggingface.co/deepseek-ai huggingface.co/deepseek-ai/...

All the hard work of the last few months has culminated in stellar performance of TGI 3.0. It's such an honor to work in this team! 🔥

Should HF do more agent stuff? If so, what would be useful?

🤗 ❤️ 🦋 (Bluesky profile on the Hub unlocked ⚡)

Huge shoutout to @dottxtai.bsky.social for making outlines-core. We just moved the first bit (JSON schema -> regex compilation) from our Python backend to Rust router. Benchmarked some schemas and it's super fast 🔥.

I had been out of #nixos for a while. The new `nix store/profile diff-closures` commands are very handy. `nix profile diff-closures --profile /nix/var/nix/profiles/system` for system rebuild changes.

Bluesky pro-tip: you can set your domain as a handle: bsky.social/about/blog/4... Bonus: you can keep your handle if you ever want to move to another server in the future.

I needed an AWQ-quantized version of Phi 3.5 MoE Instruct for testing. Might as well drop it on the Hub in case it's useful to anyone else: huggingface.co/danieldk/Phi...

I create a starter pack of current and former @explosion-ai.bsky.social. Give some of the people who made spaCy, Prodigy, Thinc, etc. a follow: go.bsky.app/VngRFva

We just added w8a8 int compressed-tensors support to Hugging Face TGI. This unlocks support for 8-bit models on GPUs with CUDA capability 7.5 (NVIDIA T4, etc.) 🔥. github.com/huggingface/...

All the bootstrapping work that Guix folks are doing is so awesome! jakstys.lt/2024/zig-rep... (Bootstraps the Zig compiler from source-only.)

Praise to the person who added color output to `ip`. So much easier to read out the different addresses. #linux

We just added initial support for compressed-tensors checkpoints to @huggingface.bsky.social text-generation-inference: github.com/huggingface/... compressed-tensors is a safetensors extension by Neural Magic to efficiently store sparse quantized tensors on disk. Support for more CT formats soon!

Slides from my PyCon LT keynote! Had some great chats afterwards with devs who felt disillusioned with ✨AI✨ & said it gave them hope, which was amazing. The AI Revolution Will Not Be Monopolized: How open-source beats economies of scale, even for LLMs speakerdeck.com/inesmontani/...

In Vilnius for PyCon Lithuania 🐍☀️ If you're at the conference tomorrow, come and say hi! I'll be giving the closing keynote titled "The AI Revolution Will Not Be Monopolized: How open-source beats economies of scale, even for LLMs". 📍https://pycon.lt/2024/talks/UGBKRS

Light Leap Years xkcd.com/2897

Out now: spacy-llm v0.7.0! 🔗 Built-in entity linking support 💬 New task for translation from/to arbitrary languages ❓ Use the Doc as prompt for question answering 🧩 Arbitrarily long docs via sharding github.com/explosion/sp...

These throwable fire extinguishers seem like a real-life example of xkcd.com/937 — they have really good reviews from everyone except the few people who tried using them to put out fires

What could possibly be a better reason for a first post on Bluesky than our lovely holiday dinner? We had a great evening yesterday, although we missed some people! Future posts will tell you more about what we do, but for now this group picture at least gives you an impression of who we are 😀

Whenever we're at conferences, people ask me where they can buy our swag to support us. So we FINALLY DID IT! Launching our new merch store, including: 📚 2 exclusive designed books 👕 supporter shirts ☕ mug, keychains, poster 💙 & of course stickers! explosion.ai/merch

📺 @honnibal.bsky.social's MLOps World talk "How many Labelled Examples do you need for a BERT-sized Model to Beat GPT-4 on Predictive Tasks?" 1. Predictive tasks still matter. 2. In-context learning isn't optimal for predictive. 3. Workflow for labelled examples. www.youtube.com/watch?v=3iax...