Hugging Face Launches Microduck: Meet the $399 Open-Source Duck Robot You Can Train With AI
TL;DR
- Hugging Face has launched Microduck, a $399 open-source duck robot designed to make AI robotics accessible for home experimentation and learning.
- The robot can be taught new skills and behaviors using reinforcement learning, with CEO Clem Delangue positioning it as a hands-on platform for AI education.
- Fully open-source and customizable, Microduck is aimed at developers, students, and hobbyists who want to build, train, and share AI-powered robotics projects.
A Hugging Face Goes From Software to Waddling Hardware
Hugging Face, the company best known as the go-to hub for open-source AI models and datasets, is diving deeper into physical AI. The company has unveiled Microduck, its latest and most charming robotics project yet — a small, bright yellow duck robot priced at $399 and built from the ground up to be hacked, trained, and loved.
Unlike closed, expensive research robots, Microduck is designed to sit on your desk and learn directly from you, bringing the kind of AI experimentation that usually requires a lab into living rooms and classrooms.
What Exactly Is Microduck?
Microduck is a compact, desktop-sized bipedal duck robot built for experimentation. While it looks like a toy with its rounded body, orange beak, and expressive eyes, it’s packed with sensors, microphones, cameras, and motors that let it see, hear, walk, waddle, and interact with its environment.
It runs on an open hardware and software stack that connects seamlessly to the Hugging Face ecosystem, including its LeRobot platform for robotics. Out of the box, Microduck can walk, respond to voice commands, follow objects, and dance — but that’s just the starting point. Every behavior is meant to be modified, improved, or completely replaced by its owner.
Why a Duck — and Why Now?
The choice of a duck isn’t just for cuteness. Hugging Face says the friendly, approachable design is intentional to lower the barrier to entry for robotics. A duck is non-intimidating, instantly recognizable, and fun to interact with, which makes it ideal for education and home use.
The launch is part of Hugging Face’s broader push into open robotics. After its work on LeRobot, open datasets for robotics, and its acquisition of Pollen Robotics earlier this year, Microduck represents its most affordable and consumer-friendly robot yet. CEO Clem Delangue has framed the project as a way to do for robotics what Hugging Face did for natural language processing: make cutting-edge tools open, affordable, and community-driven.
Trained Like a Pet, Powered by AI
The standout feature of Microduck is how you teach it. According to Delangue, owners can train Microduck to learn new tricks using reinforcement learning — the same trial-and-error technique used to train advanced AI agents.
Instead of writing complex code for every movement, users can reward the duck for desired behaviors and let the AI model figure out how to improve. Want it to learn to navigate around your desk clutter, respond to a custom hand gesture, or perform a new waddle? You can demonstrate, give feedback, and fine-tune its model locally or through Hugging Face’s tools.
The robot supports Python, integrates with popular ML libraries, and can run small vision and language models directly. That means it can be connected to open-source models from the Hugging Face Hub to give it new capabilities, from object recognition to voice interaction.
Fully Open-Source and Ready to Hack
As with all of Hugging Face’s major releases, openness is the core philosophy. Microduck’s hardware designs, software stack, training code, and model weights are open-source.
That means anyone can 3D-print replacement parts, add new sensors, modify its gait, or share their trained skills with the community. Hugging Face is encouraging users to upload their best Microduck policies and datasets back to the Hub, creating a shared library of duck skills that anyone can download and remix.
This community-driven approach is designed to accelerate learning for everyone — a beginner can download an expert’s trained model for balancing or following, then build on top of it.
Who Is Microduck For?
Hugging Face is explicitly targeting three groups: developers who want an affordable platform to test embodied AI, students and educators looking for a hands-on way to teach AI and robotics, and hobbyists and makers who just want to tinker with a cute, intelligent robot.
At $399, it significantly undercuts most research-grade robots that cost thousands of dollars, while offering far more flexibility than closed toy robots. It doesn’t require a PhD to operate, but it’s powerful enough for serious research into human-robot interaction and reinforcement learning.
Availability and the Bigger Picture
Microduck is available to order now through Hugging Face at $399, with kits expected to ship to early buyers in the coming weeks. Documentation, tutorials, and starter models are already live on the Hugging Face Hub and GitHub.
The launch signals a major shift: AI is no longer just living in the cloud. With Microduck, Hugging Face is betting that the next wave of AI enthusiasts won’t just chat with models — they’ll train them to waddle, quack, and learn right on their desks.
Get All The Latest Updates Delivered Straight To Your Inbox For Free!