Search: “clem-delangue--x”

We need more coding and agent traces public sharing to build datasets and better open source models! Lots of people contributing already, you should share yours too! https://huggingface.co/datasets?se

AI safety can't happen behind closed doors! Super cool to see that the @AISecurityInst is releasing its evals, datasets, and models in the open on @huggingface, so researchers everywhere can scrutiniz

Agree! was talking about this with @havoyan just a few days ago. That's also the reason why so much of the value has been accruing to the frontier models in my opinion (cc @GavinSBaker) because if you

Artificial intelligences do not undergo experiences, do not possess a body, do not feel joy or pain, do not mature through relationships, and do not know from within what love, work, friendship or res

Fine-tune your first AI model today. Run GPT4o level model and run on your phone or laptop. @OpenBMB released 15M samples SFT dataset that you can use right now. (319GB high-quality post training data

NVIDIA has released a corrected version of DeepSeek-V4-Pro-NVFP4, a quantized model variant optimized for NVIDIA hardware using NV-FP4 (NVIDIA's 4-bit floating-point format). The model is available on

llama.cpp now has an official website: https://llama.app Our goal is to make local AI accessible to everyone, and improving the user experience is a big part of that. On the new landing page you’ll fi

love it! @ggerganov 's llama.cpp delivering! pibot is now running fully local, using parakeet for STT, qwen3-tts for TTS, and Qwen 3.6 as the local multi-modal LLM via llama.cpp. The STT and TTS infer

Most people know Hugging Face from its public models and datasets but few realize that 50% of the models and datasets stored on HF are private. This number has been increasing with buckets (our S3 alt

Most people training agentic LLMs with RL right now have a silently broken training loop and have no idea. Here's the trap: single-turn RL works beautifully. Clean curves, sane rewards, everything con

pibot is now running fully local, using parakeet for STT, qwen3-tts for TTS, and Qwen 3.6 as the local multi-modal LLM via llama.cpp. The STT and TTS inference engines are Rust/mlx-c based. Ported fro

Thanks for the call out and glad to see that our blog inspired the broader community! We at @FireworksAI_HQ have been using delta updates to scale RL globally for about a year. Including Composer 2 &

This week, I got our GitHub Actions to use @HuggingFace Jobs instead of the default GitHub CI runners, making workflows run on much more reliable CPUs or even on serverless GPU (that cost less than a

Noah McLaughlin created a 7-billion parameter language model and published it on Hugging Face, drawing a humorous parallel to how AI practitioners evaluate models using model cards by suggesting human

This post announces an 8-9 billion parameter language model optimized for efficiency and speed, trained on a very large token dataset, and positioned as accessible for users with limited GPU resources

just noticed today - the dataset is already past 1k+ downloads. opensource / openresearch ftw ! @evo__hq would be opensourcing as many datasets, evals and autoresearch runs as we can in our pursuit of

Local Reachy Mini conversations wireless looks like magic! You can bring your friend around the house and get WOW effect from anyone! 🔥 Thanks @andimarafioti for the blog post on how to set this up: h

Narrative violation: Open model use in Factory has more than 3x’d in the last month relative to closed models. By both total consumption and event count. Will be interesting to see what open vs closed

📢 New @heyjasper release ! 📢 MONET 🌸 : An Apache2.0 deduped and recaptioned dataset of 105M samples unlocking reproducible text-to-image research. Nano T2I 🖌️ : A codebase to train your own T2I model

NVIDIA has released GLM-5.1-NVFP4, a quantized version of the GLM-5.1 model, now available on Hugging Face. The model appears to use NVFP4 (NVIDIA's floating-point 4-bit) quantization format, designed

Spectacular work by @ClementDelangue and the team at @huggingface! The HF science team just made async RL weight sync ~100x cheaper on bandwidth, and you don't need a shared cluster anymore. The probl

The HF science team just made async RL weight sync ~100x cheaper on bandwidth, and you don't need a shared cluster anymore. The problem: every RL step, the trainer typically has to sync fresh weights

today is (potentially) a great day for the GPU poors if DiffusionBlocks works on fine-tuning existing models, then literally any reasonable consumer GPU can do LLM fine-tuning will make a video on thi

Today, we're releasing LFM2.5-8B-A1B, a device-optimized model designed to power real-life applications on phones, laptops, PCs, robots, and fast & lightweight server-side use-cases. > 8B MoE, 1.5B ac

We are starting to be quite bullish about getting in the data infrastructure business. I just cloned 68 TB (while I only have a 4TB local disk) to my @huggingface training bucket in 1 minute 55 second

We're releasing Paris 2.0, which, to our knowledge, is the world's first decentralized trained video generation model. We benchmarked it against a monolithic model trained on the same data and compute

We're selectively releasing the Paris 2.0 weights and partnering with researchers and teams interested in diffusion-based video models, world models, and embodied agents. The model is on Hugging Face:

i own 0 A100s but i can run cloud inference with sub 80ms network latency by renting one in Kansas (0% action queue starvation) this is an early checkpoint of smolvla trained on 50 episodes for ~20k s

If the Founder of Hugging Face asks, you gotta do it. Models and dataset now live: https://huggingface.co/papers/2605.22391 Also built an explorer: https://huggingface.co/spaces/Kaikaku/epicure-explor

Introducing DiffusionBlocks: Block-wise Neural Network Training via Diffusion Interpretation http://pub.sakana.ai/diffusionblocks What if we didn’t have to hold an entire neural network in memory to t

jesus, qwen3-tts is FANTASTIC. going for full local stt/tts/llm with parakeet, qwen3-tts, and gemma 4 via llama.cpp for my little robot. excite, excite! https://huggingface.co/Qwen/Qwen3-TTS-12Hz-1.7B

RF-DETR is now available in @huggingface transformers state of the art in both detection and segmentation, outperforming YOLO architectures - checkpoints: https://huggingface.co/Roboflow/models - demo

Super happy that a bunch of people are finding marlin useful. Thank you for the inference support @ZeroGPU_AI @huggingface 🥰🤝 We’ve got hands on more compute now so we’ll also release a series of blog

After seeing these tweets, I decided to try it out on my own old Ubuntu computer with RTX 1070 GPU (the one that I just upgraded from 16.04 all the way to 24.04 the other day). Asked Codex on my Mac t

As Chinese AI models go closed source, and all U.S. frontier models are closed, there is a massive opportunity for a western open source AI Lab A future lack of competitive open source AI models is a

Clem Delangue advocates for the release of a 100 billion parameter Gemma 4 Mixture of Experts model, suggesting that Gemini Flash 3.5's release creates an opportunity for this larger model to be made

Gemma 4 adoption numbers outpacing Qwen 3.5/3.6 for the same sized models is a big shift in the international balance of influence via open models. Some ideas for what comes next, May 2026 Gemini Flas

Huge one for developers building with AI. Excited to announce @ClementDelangue (CEO @huggingface) is joining DASH 2026 for a fireside chat with @oliveur. Open source changed software, open models are

Introducing CHI-Bench on @huggingface: the world’s first long-horizon healthcare benchmark for AI agents. 75 real healthcare workflows + 20 apps + 200+ MCP tools + 1,290 skills + process / outcome rew

🙏 Thank you all for the incredible love and support! Our latest Tencent Hunyuan translation models are on fire on Hugging Face: 🥰Hy-MT2-1.8B ranks #1 🥰Hy-MT2-30B-A3B ranks #4 on the open-source model

Hugging Face has released 1-bit and Ternary Bonsai Image 4B, a new family of lightweight image-generation models optimized for efficient diffusion inference. These models are designed to deliver high-

Hugging Face shared insights and presentation slides detailing their process for generating 1 trillion tokens of synthetic data, offering a technical overview of their data generation methodology. The

The post reviews the Hugging Face Reachy Mini robot, praising its build quality, IDE, and developer ecosystem for desktop companion robotics applications. It highlights the robot's support for agentic

This post likely discusses the distinction between advertising and other forms of content or promotion, possibly addressing concerns about whether certain content or platforms constitute advertising.

I’ve just released MiMo V2.5-Coder. If you have 128 GB of RAM, this is one of the best models you can run locally. It’s fast, and in all my experiments it outperformed Qwen 3.6 and DeepSeek 4-Flash. h

MiniCPM5-1B is an impressive release in the 1B class! @OpenBMB https://huggingface.co/collections/openbmb/minicpm5 ✨ 1B - Apache 2.0 ✨ Hybrid reasoning with Think / No-Think modes ✨ 128K context ✨ Run