Hugging Face Microduck Open-Source Robot Release

Hugging Face unveiled Microduck, a small, open-source robot priced at $399. The robot is capable of walking, picking up objects, and getting back up after a fall. It can learn new skills through reinforcement learning using a sim-to-real (sim2real) loop, where behaviors are trained in simulation and then deployed on the physical hardware. The robot's reinforcement learning stack is open-sourced, allowing users to train it in simulation and run the learned behaviors for real. The release is positioned as lowering the barrier to entry for hands-on exploration of embodied AI and world models.
Key insights
Microduck Lowers the Financial and Technical Barrier to Embodied AI Research
At a $399 price point and with a fully open-source design, Microduck is intended to make physical AI experimentation accessible to a wider audience. The robot provides a complete sim2real loop, where users can train new skills in a simulated environment and then transfer them directly to the hardware. This addresses a major hurdle in robotics: the high cost and complexity of physical platforms. By open-sourcing the reinforcement learning stack, Hugging Face enables the community to collectively develop and share new behaviors, accelerating the exploration of embodied AI and world models.