Bertrand Chevrier

Nice to see the hardware owned by the 🤗 community

300,000 AI builders have already added their hardware to HF to instantly see what model they can run locally. To do so, go to huggingface.co/settings/local… and add your hardware specs. You can even show off publicly by adding it to your HF profile! Let's go local AI!

Kimi-K2.6 is doing good on a couple of benchmarks 🏆

Sick of those errors when compiling a GPU kernel `nvcc fatal : Unsupported gpu architecture 'compute_90'` 😡 Find pre-compiled and optimized Kernels for your hardware straight from the Hugging Face Hub. 😎

Featured Apps you can try on Hugging Face this week 🔥 🗣️ Voxtral TTS Demo: Mistral's new text-to-speech 🎙️ Cohere Multilingual ASR: multilingual transcription ⚡ Cohere WebGPU: same but locally in your browser 🎩 Mr. Chatterbox: Victorian-era gentleman chatbot 🎵 PrismAudio: text-prompted audio for video 🌍 PROMETHEUS v1.0: embodied AI world model 🎨 VFig Image2SVG: diagrams to editable SVGs 🕺 LTX 2.3 Sync: portrait animation & lipsync

I'm going to start publishing open-source datasets for TTS and STT. I'll be creating a collection for newly released models every day. Make sure to follow the Huggingface page.

ScreenSpot-Pro🤝@huggingface Our GUI computer use leaderboard is now officially live on @huggingface. Huge thanks to the HF team for the integration. The best part? The leaderboard is now community-maintained. Seeing Qwen3.5 land at #5 out of the gate is impressive, but the specialist Holo2 family still holds the crown. We're excited to see who builds the next GUI-optimized powerhouse to shake up the rankings!🔥 ScreenSpot-Pro: huggingface.co/datasets/likai…

Introducing Unsloth Studio ✨ A new open-source web UI to train and run LLMs. • Run models locally on Mac, Windows, Linux • Train 500+ models 2x faster with 70% less VRAM • Supports GGUF, vision, audio, embedding models • Auto-create datasets from PDF, CSV, DOCX • Self-healing tool calling and code execution • Compare models side by side + export to GGUF GitHub: github.com/unslothai/unsl… Blog and Guide: unsloth.ai/docs/new/studio Available now on Hugging Face, NVIDIA, Docker and Colab.

🪣 We just shipped Storage Buckets: S3-like mutable storage, cheaper & faster Git falls short for everything on high-throughput side of AI (checkpoints, processed data, agent traces, logs etc) Buckets fixes that: fast writes, overwrites, directory sync 💨 All powered by Xet dedup so successive checkpoints skip the bytes that already exist ➡️

We just shipped Community Evals and Benchmark repositories for decentralized evals 🤗 > Scores you and model authors report are on leaderboards 🙌🏻 > Benchmark datasets host live leaderboards of reported results 🚀 > You can open PRs to add scores, they live in model repositories. Community Evals will expose scores currently distributed across model cards, papers, and benchmarks. It won’t solve the differences in scores, but it is transparent!

Open Weights & Open Source v Claude Code. How does [Toad's] fractal on the title page work? Speeded up 4x. Mixture of zai-org/GLM-4.7 and openai/gpt-oss via Hugging Face Inference providers. Join Toad Explorers and get $20 of inference provider credits

HuggingChat is stupidly fast now. We made it x3 faster, I get my answer before ChatGPT even loads. + open models lowkey caught up. (x1 speed video)

Envie de faire transpirer votre service IT ? Envoyez-leur des liens générés par CreepyLink … 👉 creepylink.com

3,000 reachy minis have been shipped all over the world, just before Christmas. Happy Holidays everyone!

I’m not saying an open-source model you can run on consumer hardware is better than ChatGPT, just that sometimes it’s worth trying both. Prompt: “A zebra with a volcanic eruption in the background.”

Yep and you can follow it to watch for updates and release.

5h of Stoner, Sludge & Prog open.spotify.com/playlist/5dRMR…