ok what the fuck. this was a dry google doc a moment ago gave Opus 5.5 my Open Alignment brainstorm gdoc and asked for a video. now I want one for every doc I’ve ever written hahaha

Super happy to release SmolDataEnvs: 5,000 verifiable RL environment tasks for hill-climbing small models in code and data science by @adithya_s_k 100% open source: environments, evals, training!

Today, we open-source Pruna-Qwen-Image-2.1, a set of a few-step LoRA adapters that make Qwen-Image-2.1 by @Alibaba_Qwen up to 6.3× faster for image generation and editing. - 𝗛𝗼𝘄 𝗲𝗳𝗳𝗶𝗰𝗶𝗲𝗻𝘁 𝗶𝘀 𝗶𝘁? Generate or edit images in just 5 or 8 steps instead of 40. Choose 5 steps for maximum speed or 8 steps for the best balance. - 𝗛𝗼𝘄 𝗴𝗼𝗼𝗱 𝗶𝘀 𝗶𝘁? The 8-step adapter is our recommended default. It supports text-to-image and single- or multi-image editing at 1K resolution, with up to three reference images. - 𝗛𝗼𝘄 𝗱𝗼𝗲𝘀 𝗶𝘁 𝘄𝗼𝗿𝗸?The LoRA adapters load directly on top of Qwen-Image-2.1, while the pipeline remains unchanged. Each adapter uses its own optimized sigma schedule and runs without CFG. 𝘛𝘩𝘪𝘴 𝘪𝘴 𝘫𝘶𝘴𝘵 𝘰𝘶𝘳 𝘧𝘪𝘳𝘴𝘵 𝘳𝘦𝘭𝘦𝘢𝘴𝘦, 𝘸𝘪𝘵𝘩 𝘦𝘷𝘦𝘯 𝘣𝘦𝘵𝘵𝘦𝘳 𝘷𝘦𝘳𝘴𝘪𝘰𝘯𝘴 𝘰𝘯 𝘵𝘩𝘦 𝘸𝘢𝘺. 🚀 Want to try the fast open-source integration? Check it here on Diffusers: Want to try the fastest image generation and editing endpoint? Check P-Image-Ideogram and P-Image-Edit here on API:

Alert: Apple just dropped a new model on Hugging Face. It's a Qwen3.5-9B finetune that turns long documents into small page images to save tokens, then pulls up the full text of only the pages relevant to your question 💡

When several people talk at once, a transcript can get messy fast. Our new Nemotron 3 Diarization model tracks who spoke when, even when voices overlap. It handles up to eight speakers, has 100M parameters, and is now available on @huggingface 🤗

Introducing FLUX 3 Action. An open weights 7B World Action Model that achieves first place on the RoboLab benchmark. It outperforms the previous best open model by 6.1 percentage points while using 56% fewer parameters and running up to 3.95x faster.⁠⁠ FLUX 3 Action removes the usual trade-off between world action model performance and VLA speed: it still predicts video and actions together, but plans more than twice as far ahead and runs faster per second of robot motion than the strongest open VLA. Teams can fine-tune FLUX 3 Action on their own demonstrations to create policies for a particular robot and task. Together with @nvidia, we also integrated FLUX 3 Action natively into @huggingface's LeRobot, with fine-tuning recipes included and edge deployment on NVIDIA Jetson. Beyond robotics, we’re also seeing promising results training task-specific policies for acting in simulated environments like gaming, controlling a vehicle, computer use, and wherever else a model needs to understand a visual environment and then choose what to do next. FLUX 3 Action builds on the same image, video, and audio pretraining as FLUX 3, but uses a smaller architecture designed for practical deployment. In midtraining, we trained the model to predict actions and future frames together. We’re releasing the weights, code, fine-tuning recipe, benchmarks, and reproducible examples so researchers and developers can build on the model with their own robots, environments, and tasks (see below).

Trained a Jev-style classifier on @huggingface Jobs for ~$1.50. It's a 194M GLiNER2 model that suggests task tags for any Hub dataset from its column names and first row, and returns a label with a probability. Zero-shot, GLiNER2's first suggestion matched an owner's tag 10% of the time. After 17 minutes of fine-tuning: 69%. The fine-tuned model runs on a free CPU in about a second. Owners' tags are noisy, so some "wrong" answers are tags the owner left out. The recipe is open: one hf jobs command trains the same kind of model on your own labels. The README example (book titles) runs in ~2 minutes for about $0.02. Demo: Recipe:

We’re proud to sponsor Open Together, a free community event hosted by @huggingface on Friday, October 16, to kick off Open Source AI Week in SF. The evening is split into two parts: ➡️ 6PM – 9PM: 36 live community demos, food, drinks, and time to connect with open-source builders 🪩 9PM – 12AM: Full dance floor with live DJ sets Doors open at 6 PM, and the first 500 people through the door get collectible HF swag. 🤗 RSVP:

Voice agents still don’t understand who’s speaking to them. That’s a huge gap compared with humans, hidden by all the “phone-call” demos. But that changes today! NVIDIA is open-sourcing Nemotron 3 Diarization: a model that can reliably track speakers in live conversations, under a commercial-friendly license! In my tests, the quality is really good with one-second speech chunks. So we can use it for voice agents! I tested it with Reachy Mini and speech-to-speech running on a DGX Spark. It’s super fun to see the robot notice a new voice, ask for a name, and remember it. The model has day-zero integration with Transformers! Kudos to the NVIDIA team for shipping useful tools for the whole community!

I'm happy to announce that I've joined Hugging Face. What started as a personal project back in February is now something I get to work on full time. Local AI has grown explosively this year. I've said this since the early oMLX releases: I want my friend who bought a MacBook yesterday to be able to run AI on it today. I believe MLX has that potential. Apple Silicon is the easiest entry point for a regular person to get started with AI, with no complicated hardware to assemble. And the open source community, including Hugging Face, has been growing that potential. That support has already made a huge difference. Hugging Face is the best place for me to support oMLX and the MLX community with everything I have. For anyone getting into local AI, the first step usually starts at Hugging Face. Mine did too. I'm proud that I get to work at that entry point, where new ideas can be tried out. oMLX stays exactly where it is, under the same Apache 2.0 license in the same repository, and I'll keep leading the project, same as before. What changes is that I can now spend far more time on it, move faster, and build something sustainable for the long term together with all the contributors who have put so much into it. I also want to do more for MLX as a whole. oMLX is built on top of transformers, mlx-lm, mlx-vlm and the rest of the ecosystem, and I'm grateful to the people behind them. Rather than keeping everything inside oMLX, over time I want to push work upstream where it makes sense. And where the community needs something that doesn't exist yet, oMLX is a good place to try it first. Thank you to the 264 contributors who have built oMLX with me, and to everyone who filed issues with detailed logs and reproductions so we could fix things. oMLX would not be what it is without you. oMLX continues in the same place, in the same way. Just much faster. HF's announcement:

How do you keep @vllm_project moving at the speed of light without excluding users who run diverse models on diverse hardware? In a new PyTorch Foundation blog, contributors from @IBM, @Meta, and @huggingface introduce hardware-agnostic layers designed to balance frontier performance with portability, helping ensure vLLM continues to meet the needs of the broader open source ecosystem. Read the blog to learn more: @hmellor_ @th_ortner

Introducing the Decision Index 0.1 ⚖️ a rigorous leaderboard comparing jev with 30+ open weights decision models 35+ benchmarks. asking 130K questions to each model testing knowledge 🧠, automation ⚙️, understanding 🤔and even creativity 🎨

A great blog written by @loldedxd & @ariG23498 🤗👏 funes, by @huggingface, turns past agent sessions into memory your agents can actually use. It indexes Claude Code, Codex, pi, and Hermes traces into one local Lance dataset, then gives the agent 'recall' and 'get' tools. The next time a task depends on old reasoning, the agent can pull the original passage back. No LLM summarizing your traces at ingest.

leaving a few beginner friendly guides for classifiers (normal and zero-shot) as well as where you can find them on the Hub opt for DeBERTa and ModernBERT ones > > > multimodal (image <> text)

Qwen Image 2.1 is here! 🖼️ A 7B params native image generation and editing model, with up to 10 image references The model comes with it's own prompt enhancement LLMs, integrated with diffusers 🧨 and ComfyUI ▶️ on Spaces