Hugging Face and Humanoid Robots

Acquisition of Pollen Robotics and affordable robot arms!

I’m a big fan of Pollen Robotics, and Hugging Face also. On my recommendation, Circuit Launch built the first Reachys (v1 and 2) outside of Pollen Robotics for the pilot MechLabs education workshops. I was also a judge on the ANA Avatar XPrize where Pollen Robotics beat out some hefty university based humanoid robotics teams to take 2nd place in the finals. And I’m super excited about the impact that the Hugging Face Le Robot arm will have on physical intelligence.

I invited Remi Cadene to give a talk at the Humanoids Summit at the Computer History Museum in December and we’ve just hosted a Hackster/SeedStudios hackathon using the Le Robot arm at Circuit Launch Mountain View (see below). It’s going to be the first of many events we organize combining community and leading edge open source robotics and AI tools from Hugging Face and Pollen Robotics (and Nvidia and Open Robotics).

So you might have read the news already but I really want to share the Hugging Face blog post talking about the acquisition of Pollen Robotics!

Hugging Face to sell open-source robots thanks to Pollen Robotics acquisition 🤖

By Thomas Wolf, Clem 🤗 and Matthieu Lapeyre (READ THE ORIGINAL POST ON HUGGING FACE BLOG)

Since Hugging Face started the LeRobot library in 2024, led by ex-Tesla lead Remi Cadene, the Hugging Face Hub has quickly become the most widely used hub and software platform for open robotics with models, datasets, spaces and libraries.

Today, we’re excited to take it a step further by welcoming Pollen Robotics to Hugging Face, a team that's spent the last 9 years building open-source robots and hardware.

We believe robotics could be the next frontier unlocked by AI — and it should be open, affordable, and private. Our vision: a future where everyone in the community, from hobbyists to enterprises, can build or use robot assistants or games, starting from open solutions instead of closed, remote controlled, hardware.

This marks the fifth acquisition for Hugging Face, after notable additions like Gradio and XetHub.

According to Clem's 2025 prediction about robots, "at least 100k personal robots will be pre-ordered" this year, signaling massive growth in the space.

This month also celebrates one year since Hugging Face created LeRobot. The community of LeRobot-native DIY robot builders has flourished on YouTube and Discord, truly democratizing access to robotics technology. In just twelve months, the GitHub repository has grown from zero to over 12,000 stars, with dramatic growth continuing into 2025.

The first robot we’re offering is Reachy 2 — your friendly little lab partner for the AI era that is already in use in labs like Cornell or Carnegie Mellon. It’s a state-of-the-art humanoid robot that is open-source & VR-compatible, built for research, education, and embodied AI experiments. You can already order one for $70,000 by emailing sales@pollen-robotics.com.

Hugging Face’s Robotics Venture Timeline

2024

• March: Remi Cadene, ex-scientist on Tesla’s Optimus, joins Hugging Face to lead new open-source robotics project.
• May: Remi & HF Robotics team announce LeRobot open source robotics code library. Becomes largest and most comprehensive open-source library for robotics, with over 12k stars on GH and over 100 repositories on the HF Hub.
• June: Hugging Face and Pollen Robotics collaborate on Reachy 2, the first open-source robot powered by LeRobot
• August: Robotics team releases in-depth tutorial for developers to build and train their own AI-powered robots at low cost
• October: LeRobot partnered with @therobotstudio to release the SO-100 arm — a high-performance robotic arm for around $100. It’s the most accessible option out there, and one of the easiest DIY builds you can do at home!
• November: NVIDIA announces collaboration with HF to accelerate LeRobot’s data collection, training, and verification workflow

2025

• March: NVIDIA announced GR00T N1, released on Hugging Face — the world’s first open foundation model for humanoid robots — and fine-tuned it to run on the LeRobot SO-100 robotic arm.
• March: LeRobot teamed up with Yaak to release Learning to Drive (L2D) to the robotics AI community. L2D is the world’s largest multimodal dataset aimed at building an open-sourced spatial intelligence

About Hugging Face

Hugging Face is the most used platform for AI builders, with over 7 million users and a thriving community of researchers, developers, and organizations. Founded by Clément Delangue (CEO), Julien Chaumond (CTO), and Thomas Wolf (CSO), the company has become a central hub for AI, hosting the most popular open-source models and datasets, including Meta’s LLaMA 4, DeepSeek R1, Stable Diffusion, Black Forest Flux, Mistral, OpenAI Whisper and millions of others. Its AI app store features over 500,000 ready-to-use applications called spaces, making state-of-the-art AI accessible to all. Hugging Face has raised over $395 million to date from leading investors including Google, Amazon, Nvidia, Salesforce, Sequoia, Coatue, Lux Capital, and Addition.

About Pollen Robotics

Pollen Robotics, established in 2016, is a leading developer of open-source humanoid robots designed for advanced research and real-world applications. Founded by former researchers from Inria's Flowers team, including Matthieu Lapeyre and Pierre Rouanet, the company leverages years of experience at the intersection of robotics hardware and applied artificial intelligence, to develop innovative and user-friendly robotic platforms. Reachy 2 is the latest generation of Pollen Robotics’ flagship humanoid robot. Designed as a versatile platform for embodied AI, Reachy 2 combines advanced hardware with user-friendly software, enabling intuitive interactions and agile manipulation. With open-source hardware and software at its core, Reachy 2 provides researchers, and developers unprecedented freedom to explore human-robot interactions, machine learning applications, and practical AI-driven tasks.

Reachy has been adopted by esteemed institutions such as Accenture, CEA, CNRS, Cornell University, and Carnegie Mellon University, underscoring its versatility in fields ranging from advanced AI research to practical manipulation tasks. Pollen Robotics has won several awards including second place at the prestigious ANA Avatar XPRIZE in 2022, highlighting its innovative prowess in remote presence, dexterous manipulation and control technologies.

With hundreds of units deployed across more than 20 countries, Pollen Robotics has redefined the capabilities of open-source humanoid robots, fostering collaboration and customization in the global robotics community.

About Reachy 2

• Unique, friendly, and approachable design that invites natural interaction, making Reachy instantly engaging and accessible for users of all backgrounds. A viral innovation of Reachy's design is the Orbita joint system, which provides smooth, multi-directional movement for the robot's neck and wrists, enhancing its expressiveness and interaction capabilities.
• Human-inspired arms (7-Dofs) and unique Orbita joints allow expressive, multi-directional movement, enhancing human-like interactions, with the capability to manipulate object up to 3kg
• A mobile base equipped with omniwheels and LiDAR makes navigation seamless,
• Its VR teleoperation enables intuitive remote presence, literally seeing through the robot’s eyes.
• Reachy 2’s system offers everything you need for machine learning, delivering strong performance and seamless compatibility with modern AI frameworks.
• Reachy’s open-source nature is designed to foster collaboration and customization. Pollen Robotics offers comprehensive resources, including software, 3D models, and documentation, available on their Hugging Face organization: https://huggingface.co/pollen-robotics
• Adopted by prominent research labs like Cornell University or Carnegie Mellon University.

READ THE ORIGINAL POST ON HUGGING FACE BLOG

Hackathon Features Robots Powered by NVIDIA Isaac GR00T N1 🔗

The Seeed Studio Embodied AI Hackathon, which took place last month, brought together the robotics community to showcase innovative projects using the LeRobot SO-100ARM motor kit.

The event highlighted how robot learning is advancing AI-driven robotics, with teams successfully integrating the NVIDIA Isaac GR00T N1 model to speed humanoid robot development. A notable project involved developing leader-follower robot pairs capable of learning pick-and-place tasks by post-training robot foundation models on real-world demonstration data.

How the project worked:

• Real-World Imitation Learning: Robots observe and mimic human-led demonstrations, recorded through Arducam vision systems and an external camera.
• Post-Training Pipeline: Captured data is structured into a modality.json dataset for efficient GPU-based training with GR00T N1.
• Bimanual Manipulation: The model is optimized for controlling two robotic arms simultaneously, enhancing cooperative skills.

The dataset is now publicly available on Hugging Face, with implementation details on GitHub.

Learn more about the project.

Embodied AI Hackathon at Circuit Launch Mountain View is ON! Participants will team up to build 3D-printed robot arms using the Hugging Face hashtag#LeRobot platform—the hands-on experiment of imitation learning with ACT/Diffusion policy/Pi0 training and deployment on Jetson Orin. 🤗💚 Thanks to ecosystem support! NVIDIA Robotics, Hugging Face, Hackster.io, Circuit Launch! Happy Hacking!

Want to be part of more amazing events?

Join the organizing team at Silicon Valley Robotics - bots&beer@svrobo.org