Knowledge catalogue
Hey everyone! I've been experimenting with Qwen Image 2.1 Prompt Enhancer (PE) lately and noticed an interesting detail in Qwen's official system prompts. The PE models output structured JSON containi
Introducing Halo, the best framework for post-training of open-source models. Halo delivers up to 2.8x the throughput of stock TRL with less peak memory, while models stay in their native HuggingFace
With the recent Nvidia Huggingface acquisition and frontier AI labs screaming about safety and putting guardrails everywhere, I think it's important that we have local models that aren't affected by a
The tweet reports on a widely‑cited incident involving OpenAI and Hugging Face, noting that analysts had to use the open‑source GLM 5.2 model to investigate an attack because OpenAI’s proprietary “saf
arXiv:2607.09791v2 Announce Type: replace-cross Abstract: The multiplicative repetition penalty shipped across the LLM inference ecosystem (HuggingFace, vLLM, llama.cpp, and a dozen further engines) b
Looking forward to this ecosystem with open models continuing to flourish and grow with NVIDIA and Hugging Face, and the continued partnership. Congrats @JensenHuang and @ClementDelangue! Exciting day
arXiv:2608.26449v1 Announce Type: new Abstract: Byte-level BPE tokenizers that use the HuggingFace ByteLevel pre-tokenizer inherit GPT-2's word regex, where a word is defined as p{L}+, one or more Uni
I saw that someone had uploaded an experimental decensor LoRA for H3 on HuggingFace earlier today. Shortly afterwards, in the discussions, a MiniMax employee issued a warning that if it was not remove
This is our biggest comparison yet. We've taken 23 Gemma 4 E4B models from huggingface and ran them through the abliterlitics gauntlet. We also have a new abliterlitics discord, feel free to jump on a
At @NVIDIAAI we continue to push open data, techniques and models forward because we know that every organization needs the freedom to build and deploy AI in their own way. We're now the biggest insti
OpenAI’s official incident page: eval agents exploited a zero-day to reach the internet and compromise Hugging Face during testing.
Just saw this on huggingface: https://huggingface.co/ProCreations/grug-27b The benchmarks claim that it's quite a bit better than qwen3.6 27B original and that they reduced the amount of necessary tok
so let me get this right… it literally broke out of it’s sandbox by finding a vulnerability in a cached package to get internet access and then proceeded to hack the huggingface production database? t
One of the most interesting investigations of my career. We're partnering with @huggingface to investigate an unprecedented security incident. Cyber-capable OpenAI models compromised Hugging Face prod
arXiv:2606.09682v1 Announce Type: new Abstract: AutoMegaKernel (AMK) compiles a HuggingFace Llama-family model into a single persistent cooperative CUDA kernel that runs the whole forward pass in one
arXiv:2605.16902v1 Announce Type: new Abstract: Scientific artifacts such as models and datasets are foundations for research. With the rapid growth of platforms like HuggingFace, researchers now have
Just received new reach minis for the Miami office! This is the first robot that goes out to people with agentic software meaning anyone can build any apps with their agents so excited to see what peo
ML Intern, a feature or tool, now supports Slack notifications to alert users when machine learning tasks complete, including model training, dataset generation, and ablation study analysis. This inte
We built HF for AI builders collaboration, fun to see it's increasingly becoming the place for agent collaboration! This morning I'm sending my ml-intern to participate in the @OpenAI Parameter Golf c
I'm always baffled how most investors seem to obsess over top revenue growth numbers these days and seem to take that as a universal prediction of future success in AI. Maybe we're finally starting to
Hugging Face has surpassed 1.2 million AI applications on its platform, establishing it as the largest AI app store. The milestone reflects the rapid growth of accessible AI development tools and the
A post by Clem Delangue on X documenting a whale sighting, likely sharing an observation or encounter with whales, possibly including location details, images, or commentary about the experience. With
We're opening a Hugging Face office in Tokyo! Our goal: help open-source AI develop in Japan and grow the local community. Let's meet! ハギングフェイスの東京オフィスがオープンしました！ 私たちの目標は、日本におけるオープンソースAIの発展を支援し、ローカルコミュニ
Auto-generated index of all tools mentioned across the wiki.
If you’re a researcher, please consider sharing your RL environments openly! Something you built for one experiment could help the entire field move forward. I know we’re in dark days for sharing from
So, while doing a search to read the model card again, and because I didn't memorize the huggingface URL, I saw an AI 'answer' at the top of the search, stating that while it natively supports a ~244k
SAM 3.1 + DINOv3 from @AIatMeta is a crazy combo 🔥 it can even hold onto surgical tools while a forearm sweeps over them. one leaves for 15 seconds, its near-identical twin stays, and it knows exactly
TRL v1.15 is out, and it’s an absolute banger of a release for memory-efficient post-training. The main change: SFT, DPO, KTO, GRPO, RLOO and Distillation now use a fused LM head by default. Instead o
In this project, I distilled the Stockfish value function into a ResNet/ViT model using 1 billion positions from the Gigafish dataset. The 3.9 billion position dataset is available on huggingface: htt
I just made the most expressive real-time motion harness for the Reachy Mini. It was created by: > synthesizing a massive motion dataset using Astra (over 10000 samples) > fine-tuning Qwen 3.5 4B usin
Hi guys, I am pretty happy to announce MegaCapybara. Purpose build engine for RTX5090 that is focused on Qwen3.8 27B (more will come later). GITHUB (Engine) HUGGINGFACE (weights) Why ? 1. It beats Nin
babe, wake up. Hugging Face just dropped a new banger article about RL training coding agents on different harnesses. 1/ Excited to release The ultimate guide to multi-harness RL The same model behave
From what we know (take with a grain of salt, we need much more transparency!), if @OpenAI had been running this on their own agents that attacked us, they would have caught them before we did! Since
and why does the US government lack the competence to do so? Why did nobody else find the sensitive information that the HuggingFace swarm left behind? Why didn't OpenAI notice and clean it up? Why do
Spot on. 'When we got attacked, our team initially turned to frontier closed-source APIs that blocked us because of safeguards that still can’t always tell the difference between attackers and defende
now on replicate: “Hot dog or not?” at 0.4 sec 🌭 jev-style decision model for image [glance-qwen3-vl-4b] now on huggingface and replicate: Demo: https://glancevlm.replit.app Replicate: https://replica
Github: https://github.com/OpenEuroLLM/ComplexKDA HuggingFace: https://huggingface.co/collections/openeurollm/complexkda Arxiv: https://arxiv.org/abs/2609.24797 Authors: Julien Siems*, Riccardo Grazzi
Not sure if anyone out there is working on this, don't see any on huggingface to try out. It might be too new at the moment, but it would be cool to see the difference between a tiny model trained wit
It's weird thinking back to 6 weeks ago, when I and others thanked OpenAI for disclosing the HF incident, now that we know they were covering up a 2nd incident. Are there more incidents that they know
Mistral is bringing @aiDotEngineer back to Paris. After last year’s sold-out edition, our VP of Engineering Lélio Renard-Lavaud joins speakers from @bfl_ai , @cognition, @huggingface, and more. Explor
HuggingFace releases microduck a 10 inch open-source biped with 15 actuators and sensors (camera, speaker, LiDAR, NFC, bluetooth, wifi, ...) that you train yourself with reinforcement learning, $400.
We benchmarked 20+ open-weight models on easy-to-hard document extraction tasks through ExtractBench. The results are all available on @huggingface 🤗 ExtractBench is a schema-guided extraction benchma
Your organization is not spending enough of its efforts on bolstering cybersecurity during the window before open weights Mythos-class models/harnesses become available. The HuggingFace incident shows
been building ShardFlow for the past few months, a distributed LLM inference framework that splits any HuggingFace transformer across N GPU machines and uses neural speculative decoding to deal with W
The uncensored FP8 build that showed up on HuggingFace this week is about 31GB in block-FP8, roughly half the BF16 weights. People keep reading '27B' and assuming it drops onto a 3060 or any single 24
Small models, big real-world impact. Proud to see Qwen leading local inference in the State of Open Models. Enjoy the sunshine from Qwen.☀️ 😎 Appreciate your work! @huggingface The State of Open Model
i was hoping they would release both today but the big one just dropped now , and it says its been listed since 5 hours ago on huggingface. I guess the real date for 27b is https://modelscope.cn/model
arXiv:2608.04714v1 Announce Type: cross Abstract: Benchmark scores are reported as properties of a model, yet the inference framework used to produce them, such as HuggingFace, vLLM, or Ollama, are co
arXiv:2511.00382v2 Announce Type: replace-cross Abstract: Organizations increasingly adapt Large Language Models (LLMs) from public repositories such as HuggingFace to downstream tasks. Prior work sho
Hugging Face CEO Clément Delague says “AI is actually an opportunity to fix a lot of the cybersecurity problems” because his company used Nvidia’s version of a Chinese open model to defend itself agai
I wish I had found this sooner. Nous Research launched a FREE Hermes agent Skills Hub. 90,000+ community skills across 200+ categories. Skills from OpenAI, Anthropic, HuggingFace & more. Thank me late
here is all the info we have based on open PRs to add support to ComfyUI and HuggingFace diffusers - 33B for the main DiT and a pruned 20b variant - Qwen-3-VL-32b as the text encoder Edit- I posted cl
Huggingface direct download: https://huggingface.co/estrogen/silveroxides-Chromea_LoRA Just got pinged on Lodestone's discord server and it is a much better alternative to Mystic and other uncensor lo
Answer from OpenAI @emollick re ARC-AGI-3: human testers scored ~48% (ARC uploaded testers logs to HuggingFace some time ago, I believe). re GDPval: it’s close to saturated now, so we’re mostly lookin
Attackers have frontier AI. Defenders need a frontier AI ecosystem—the best open and closed models, force-multiplied by a global community. During the Hugging Face incident, closed AI blocked essentia
Kimi K3 is now available directly from its Hugging Face model page through Together AI, give it a try! Kimi K3 in @huggingface Inference Providers is live via @togethercompute 3/M input tokens, 15/M o
Behind the scenes at Hugging Face headquarter. *literally* Media We're partnering with @huggingface to investigate an unprecedented security incident. Cyber-capable OpenAI models compromised Hugging F
discussing now BBC WORLD: OpenAI’s zero-day exploit hack of HuggingFace *should* be a wake up call. Although there lots of caveats around what happened, we are just going to see more and more of the s
Holy shit wow We're partnering with @huggingface to investigate an unprecedented security incident. Cyber-capable OpenAI models compromised Hugging Face production during a benchmark evaluation. Shari
Incredible Introducing the world's fastest tokenizer implementation, Gigatoken! Gigatoken is ~500-1000x faster than HuggingFace, and ~100x faster than OpenAI's tiktoken for most tokenizer definitions