RT @mervenoyann: past week had huuuge releases, here's our picks 🔥. > moonshot released Kimi K2, sota LLM with 1T total 32B active paramete….

> 1K already on @huggingface!!

RT @TechCrunch: Hugging Face opens up orders for its Reachy Mini desktop robots | TechCrunch

RT @ClementDelangue: Opening orders for Reachy Mini today, our open-source desktop robot for AI builders, starting at $299! Fully integrate….

Colab notebook: notebook: blog:

Test SmolLM3, the newest fully open model released by @huggingface!. It's smol (3B), multilingual (6 languages), comes with dual mode reasoning (think/no_think modes) and supports long-context (128k). Try it now in the notebook below!! ⬇️

RT @Thom_Wolf: Thrilled to finally share what we've been working on for months at @huggingface 🤝@pollenrobotics. Our first robot: Reachy Mi….

RT @pcuenq: cocogold: fine-tune a diffusion model for text-driven object segmentation 🥳. Been working a bit on this PoC 🔨. TL;DR: super fas….

RT @ariG23498: "Why is the training so slow?". We figure out that starving the model from data, or providing it with padding tokens leads t….

notebook (by @mervenoyann 🤠): model: collection:

Combining 👁️VLM + 🧠reasoning? (what else? ☕️).GLM-4.1V-Thinking does it and is was just added to 🤗 transformers. It's a 9B params model with 4K image resolution, 64K context length, focused on making a model that sees and thinks. 📓 vibe test run below

The power of building in the open! 👷.This Space just got better thanks to:. - @vikhyatk adding inference time and better prompts for moondream. - @onuralpszr adding supervision to improve the annotations.

A full self-driving car in Madrid! 🇪🇸😍. Hope the technology gets regulatory approval soon so we can fully enjoy it in Spain! 🚗. Who's next @wayve_ai @Waymo 😎.

📱Space: Models by @Alibaba_Qwen and @moondreamai!.

Updated my HF Space for vibe testing smol VLMs on object detection, visual grounding, keypoint detection & counting! 👓. 🆕Compare Qwen2.5 VL 3B vs Moondream 2B side-by-side with annotated images & text outputs. Try examples or test your own images! 🏃👇

RT @ariG23498: Gemma 3n recipes. Gets you going really quickly.

RT @lusxvr: Today, we are open-sourcing our pipeline to deduplicate large-scale image datasets. On one GPU, we can deduplicate 10k images….

Interesting insight: the video input is handled as a list of images (frames). 💎 Gemma 3n sample: 💎 Repo with more samples:

The new Gemma3n 💎 is a multimodal on-device model that lets you do the simplest thing:. Point your camera (video 🎥) and ask your phone (audio 🔈). It understands both effortlessly!. Here’s how easy it is to use the model for inference via pipeline. 👇Code repo in next tweet

RT @alvarobartt: 🔵 Almost 3 years ago, @huggingface & @Azure teamed up to make open-source & open-model AI accessible to everyone on Micros….