Hugging Face @huggingface

The AI community building the future. https://t.co/TpiXQMQ9rZ huggingface.co NYC and Paris and 🌏 Joined September 2016

Tweets

14K
Followers

708K
Following

222
Likes

9K

Sergio Paniego @SergioPaniego

17 hours ago

continuous batching just landed in TRL for GRPO at 64 generations it runs faster and uses less VRAM than plain generate, no vLLM needed how it works and when to reach for it, below

Sergio Paniego @SergioPaniego

19 hours ago

x.com/i/article/2067…

1 4 47 20K 57

4 6 46 17K 47

View Details

Can a VLM see without a vision encoder? We trained one for $100, inspired by Gemma 4 12B. Latency on an M3 Pro MacBook: 112 ms -> 1.1 ms for the image path 30% lower end-to-end image+LLM The architecture is just: patchify the image -> linear projection with pos embeddings -> LLM Writeup: huggingface.co/spaces/Hugging…

16 97 662 54K 618

View Details

Wauplin @Wauplin

2 days ago

`hf upload` got a full rewrite! Single-pass hashing, multi commits, resumable uploads. Same CLI, way faster, way cleaner. Available in hf latest release (1.20.0).

4 9 58 15K 5

View Details

clem 🤗 @ClementDelangue

a day ago

clem 🤗 @ClementDelangue

a week ago

Decided to go to DC next week to talk directly with policymakers. Not sure how impactful it will be but with everything happening, feels like a good time to share more about open-source AI, transparency, concentration of power, the real risks vs the real benefits. Who do you

62 47 770 119K 57

12 10 147 27K 9

View Details

Poolside @poolsideai

2 days ago

Today we’re releasing the weights for Laguna M.1, our most capable model to date, with a 256K context length. Both base and post-trained checkpoints are now available on Hugging Face under Apache 2.0.

40 117 1K 605K 579

View Details

Victor M @victormustar

2 days ago

Open source MUST win 🔥 GLM-5.2 is free when used with Hugging Face Inference Providers and for every available provider for the next 6 hours (Zai, Together AI, Novita, Fireworks, DeepInfra) the cost is on us. Set it up with Pi, opencode, Codex, Claude Code or any coding agent to understand why people are saying open source has caught up 🔥

59 57 608 234K 240

View Details

Liquid AI @liquidai

2 days ago

Introducing LFM2.5-Embedding-350M and LFM2.5-ColBERT-350M: two multilingual retrieval models built for ultra-fast and accurate search across 11 languages. > End-to-end retrieval latency as low as 1.5ms with our enterprise stack! 🚀 > Consistently best-in-class multilingual and cross-lingual performance across Arabic, German, English, Spanish, French, Italian, Japanese, Korean, Norwegian, Portuguese, and Swedish. 🧵

28 137 899 94K 463

View Details

MOSI @MosiAI_Official

2 days ago

MOSS-TTS Local Transformer v1.5 is here. Clone any voice. Speak any language. Hear every detail. 30+ languages, 48 kHz clarity.

15 38 366 84K 434

View Details

Eric Newcomer @EricNewcomer

3 days ago

Soaring Costs Prompt Fresh Interest in Open Source AI. Chinese Firms Are Way Ahead. newcomer.co/p/soaring-cost…

7 19 56 15K 15

View Details

Xenova @xenovacom

3 days ago

Before Fable 5 was shut down, it pushed Gemma 4 to 255 tok/s on WebGPU. Some didn't believe it was real. Today we're releasing the demo and kernels it wrote for you to see yourself. Run it locally in your browser. Agentic kernel optimization is the future of on-device inference

Xenova @xenovacom

a week ago

I gave Fable 5 one job: write custom WebGPU kernels for Gemma 4 inference. It climbed to 84 tok/s, then hit a wall, insisting further optimization was impossible. Hours later, Anthropic rolled back invisible LLM development safeguards, and it hit 255 tok/s. The next day,

145 369 5K 1.1M 2K

78 207 2K 341K 2K

View Details

Omar Sanseviero @osanseviero

2 days ago

Gemma 4 with 255 tokens per second running directly in your browser x.com/xenovacom/stat…

Xenova @xenovacom

3 days ago

78 207 2K 341K 2K

5 11 151 25K 74

View Details

Omar Sanseviero @osanseviero

2 days ago

Stupidly fast generation is all you need

32 8 228 28K 14

View Details

Julien Chaumond @julien_c

3 days ago

Llama.cpp has a new branding + official website. Run local models today! Now more than ever, open source must win. 🙏 By @alekgrygier and @ggerganov at ggml/hf

14 27 229 17K 52

View Details

Ai2 @allen_ai

3 days ago

We're releasing MolmoMotion, a 3D motion forecasting model. Given one or a few video frames, 3D points on an object, & an instruction like "Put the white bowl on the table," MolmoMotion predicts where those points will go over the next few seconds in a shared 3D world frame. 🧵

11 62 363 183K 211

View Details

Adithya S K @adithya_s_k

3 days ago

You can now train on 350+ RL Environments from OpenReward with TRL with just a few lines of code

Sergio Paniego @SergioPaniego

3 days ago

x.com/i/article/2067…

3 3 34 23K 38

2 15 151 21K 98

View Details

John Sarihan @jsarihan

3 days ago

Today, we're launching Crosby Intelligence to push the frontier of legal AI forward.

45 39 303 161K 123

View Details

John Sarihan @jsarihan

3 days ago

We’re announcing 3 things as part of the Crosby Intelligence launch today: 1/ RedlineBench with @micro1_ai – publishing the first benchmark measuring how frontier models handle multiple steps of a complex, real world contract negotiation*, hosted on @huggingface 2/ The Crosby Intelligence Research Fellowship – funding two fellows pursuing frontier research with support from @OpenAI: $25K + $12.5K Codex credits each 3/ Hosting the most interesting conversations in applied AI at our Soho office, featuring @paraga, @rahulgs, @PeterHndrsn, @NeelGuha, and more *built externally with no client data Read more about all three at intelligence.crosby.ai

4 8 65 11K 26

View Details

Pieter Delobelle @pieterdelobelle

3 days ago

Today we release Nemotron-Personas-Belgium in collaboration with @NVIDIAAI , a synthetic dataset of 4x300k Belgian personas in Dutch, French, German and English. The dataset is modeled on the @Statbel_en census, modeling age, occupations, household, income, names etc.

7 6 43 9K 14

View Details

clem 🤗 @ClementDelangue

3 days ago

Very cool and massive dataset on @huggingface of course: huggingface.co/datasets/XDOF/…!

Philipp Wu @philippswu

3 days ago

Announcing @xdofai: We’ve raised $70 million to build the core robotic infrastructure ecosystem for robot foundation models. My cofounders Fred (@YideShentu), Nemo (@itsnemojin) and I have been pursuing the dream of general purpose robots for our entire lives. After work at

86 71 687 163K 279

11 14 76 18K 21

View Details

Sayak Paul @RisingSayak

3 days ago

A quick `hf sync` a day keeps the agents at bay 😉 After you are done with your Claude (or whatever agent) session, just run a quick `hf sync` to never lose those valuable traces. You can also sync them back on a different machine. Get to work!