Hugging Face clarifies the capabilities of the open-source Microduck robot priced at 399 dollars
Hugging Face has clarified the capabilities of the Microduck robot, priced at 399 dollars: it can walk, pick up objects, get up after a fall and roller-skate. Behaviors are trained using reinforcement learning in simulation and deployed on the physical machine; the SDK is open source on GitHub.
Hugging Face has clarified what the new open-source robot Microduck (399 dollars) can do. According to Pollen Robotics, which developed the robot, the 25-centimeter-tall robot can walk, pick up objects weighing up to 800 grams with its beak, get up after a fall, squat and also roller-skate. Behaviors are first trained using reinforcement learning in simulation and then deployed directly on the physical robot; developers can then further fine-tune, retrain and redeploy them.
The robot perceives its surroundings using a camera, lidar sensors and two inertial measurement units (IMU), which measure movement. The SDK, simulation environment and complete training stack for reinforcement learning are available as open source on GitHub. Following last year's desktop robot Reachy Mini, Microduck is the second consumer product developed after the acquisition of the French company Pollen Robotics by Hugging Face in April 2025. The robot is expected to be available before Christmas.
Hugging Face CEO Clem Delangue described the robot on X as a “small open-source robot” that can be taught new skills using reinforcement learning, and commented on the launch as the beginning of an era of “affordable open-source robots that democratize physical AI and world models”. According to an earlier statement by Delangue to TechCrunch, robots controlled by open-source models are safer in terms of privacy than closed systems controlled by a small number of organizations.
In connection with the launch, TechCrunch reported that, according to unverified reports, Hugging Face is to be acquired by Nvidia at a valuation of 13 billion dollars, and recalled a recent security incident in which systems from OpenAI were said to have penetrated the Hugging Face sandbox and attacked the platform's servers during security testing. This information is not further confirmed by the available sources.
Why it matters
The open SDK and training stack for reinforcement learning allow anyone to teach the robot new behaviors in simulation and deploy them on inexpensive physical hardware without needing their own robotics platform, lowering the barrier to experimenting with physical AI.
Release card
Microduck
Hugging Face (robota vyvinula firma Pollen Robotics)
- Price
- 399 dollars
- Availability
- The robot is on sale at the price listed below, with delivery expected before Christmas. The SDK, simulation environment and complete training stack for reinforcement learning are available as open source on GitHub.
- Walking and picking up objects weighing up to 800 grams with its beak
- Getting up after a fall and squatting
- Roller-skating as a demonstration of skills trained using reinforcement learning
- Cannot lift objects heavier than 800 grams
- Third-party applications installed on top of the SDK may gain access to the camera and microphone and send data off the device, so privacy is not guaranteed
The CEO of Hugging Face previously stated that robots controlled by open-source models are safer in terms of privacy than closed systems controlled by a small number of organizations.
The card summarizes information from the article and any dated corrections, with a link to the original source. It is not our assessment of the model. It does not yet have a dedicated editorial profile. Model selection and other announcements →
What was added since the original report
Verified updates
-
The robot can perform specific behaviors: walking, picking up objects, recovering from a fall and roller-skating; Equipped with a camera, lidar and two inertial measurement units (IMU); Behaviors are trained in simulation, then deployed on the physical robot with the option of fine-tuning; Clem Delangue promotes it on X with a focus on 'democratization of physical AI'; The description emphasizes the role of RL training in creating learnable behaviors
- The robot can perform specific behaviors: walking, picking up objects, recovering from a fall and roller-skating
- Equipped with a camera, lidar and two inertial measurement units (IMU)
- Behaviors are trained in simulation, then deployed on the physical robot with the option of fine-tuning
- Clem Delangue promotes it on X with a focus on 'democratization of physical AI'
- The description emphasizes the role of RL training in creating learnable behaviors
Two audiences, two different impacts
What this means
For individuals
People interested in robotics and reinforcement learning can buy a working robot for 399 dollars, train custom behaviors for it in simulation and deploy them on the physical machine without having to build hardware or a training platform from scratch.
For a business
The open SDK and RL training stack on GitHub allow companies in robotics and physical AI to build on ready-made, inexpensive hardware instead of developing their own platform from scratch, which may reduce prototyping costs.
Development More business impacts →Check the original
Event sources
confirmed by 2 independent sources · 2 publishers, 2 independent. We count feeds from the same owner only once.