Hugging Face has introduced Microduck, a compact bipedal robot designed to make physical AI and reinforcement learning more accessible. Developed by Pollen Robotics, the 25-centimeter robot is now available for pre-order at an introductory price of $399, before taxes and shipping. First deliveries are targeted for before Christmas 2026.
The launch extends Hugging Face’s push into affordable open-source robotics. Earlier, the company brought Pollen Robotics into its organization and launched Reachy Mini as a platform for human-robot interaction. However, Microduck focuses more directly on movement, physical AI and learning through simulation.
Microduck Brings Reinforcement Learning Home
Microduck measures 25 centimeters and weighs less than 800 grams. It includes 15 motors, a camera, LiDAR, two inertial measurement units and an articulated beak for grasping objects. Moreover, the robot includes a game controller and comes with trained behaviors ready to use.
Out of the box, Microduck can walk, sit, crouch and recover from many common falls. It can also kick objects, pick up items, and roller skate when equipped with the appropriate accessories. Therefore, users can explore physical AI without first building a robot from individual components.
The more significant feature, however, is its reinforcement-learning workflow. Developers can train behaviors inside a physics simulation before transferring them to the physical robot. They can then adjust the simulation, retrain the policy and deploy it again.
Open Software Supports Custom Robot Behaviors
Pollen Robotics has made the Microduck SDK, simulation environment and reinforcement-learning stack available as open-source software. Consequently, developers can inspect, modify and retrain the behaviors running on the robot. The company says seven trained moves ship with the platform.
The robot uses a Rockchip RK3566 processor with an AI accelerator, alongside 1GB of RAM and 32GB of storage. It also supports Wi-Fi and Bluetooth, and its removable battery provides roughly 1 hour of operation, depending on usage.
Hugging Face CEO Clem Delangue said: “Welcome to the era of open-source affordable robots to democratize physical AI and world models!”
A Small Platform For Physical AI
Microduck targets developers, educators, makers and robotics enthusiasts who want hands-on experience with physical AI. Its compact size also makes repeated experimentation easier, since failed behaviors can usually end with the robot simply falling over.
Meanwhile, the $399 price point puts a programmable reinforcement-learning platform within the reach of consumer and classroom budgets. However, accessories cost extra, including a charger pack and development pack.
Overall, Microduck represents Hugging Face’s broader effort to bring open AI principles into physical machines. Instead of keeping robotics development inside specialized laboratories, the company is giving users a small platform for training, testing and sharing new behaviors.








