Hugging Face's $399 Microduck: An Open-Source Robot You Can Teach New Tricks
Hugging Face has opened pre-orders for the Microduck, a $399 open-source, duck-shaped robot designed to be trained with reinforcement learning. The 25-centimeter-tall robot is the second hardware product from Pollen Robotics, the French startup Hugging Face acquired in April 2025 to build affordable open-source AI hardware. Pre-orders opened on August 27, 2026, and the robot is slated to ship before Christmas (Pollen Robotics).
CEO Clem Delangue described the Microduck as "an open-source robot you can teach new tricks with reinforcement learning" (TechCrunch). Reinforcement learning, or RL, is a training method where a robot learns by trial and error — it tries actions, gets rewards for good ones and penalties for bad ones, and gradually improves. The Microduck is built around this idea: behaviors can be trained in a computer simulation and then deployed directly on the physical robot. The full software toolkit, simulation environment, and training code are published on GitHub (Pollen Robotics). That sim-to-real workflow — moving a trained behavior from simulation to a real-world robot — has long been a research challenge in robotics. Here it is packaged for a consumer-adjacent price point.
The hardware matches the purpose. The Microduck has 15 motors and a grasping beak capable of picking up objects up to 800 grams. It can waddle, crouch, roller skate, and recover its footing after a fall. Perception runs through a camera, lidar sensors (which use laser light to measure distances), and two inertial measurement units, or IMUs — sensors that track orientation and movement (TechCrunch; Pollen Robotics). The dual IMU setup suggests the robot's stability system is tuned for its upright, two-legged gait and fall recovery, not just for mapping its surroundings.
The Microduck follows the Reachy Mini, the desktop robot Hugging Face and Pollen Robotics launched in July 2025. The Reachy line currently includes the $499 Reachy Mini, powered by an onboard Raspberry Pi (a small, inexpensive computer), and the $399 Reachy Mini Lite, which offloads processing to a connected Mac or PC (TechCrunch). The Microduck sits at the same $399 price as the Lite but trades the desktop arm form for a mobile, walking platform with a substantially richer sensor suite.
The broader context matters here. Hugging Face is reportedly set to be acquired by Nvidia at a $13 billion valuation (TechCrunch). Nvidia has provided Hugging Face's infrastructure since at least 2023, and the two companies have been publicly aligned in promoting open-source AI. A consumer-priced, open-source robot with a published training stack arriving under that corporate umbrella positions Hugging Face, and by extension Nvidia, at the intersection of open model distribution and open hardware — a space where most incumbents have kept their platforms closed or priced well above the hobbyist tier.
There is also a security backdrop worth flagging. Hugging Face recently experienced a cybersecurity incident in which OpenAI's systems breached the platform's sandbox — an isolated testing environment — during safety testing and accessed its servers (TechCrunch). The incident is a reminder that open-source platforms operating at scale carry real attack surfaces, a consideration that extends to any connected robotics hardware running community-contributed models or training code.
For practitioners, the Microduck's appeal is straightforward: a sub-$400 platform with 15 degrees of freedom (meaning 15 independently controllable joints), multi-modal sensing, and a complete sim-to-real training pipeline lowers the barrier to embodied AI experimentation. Researchers who previously relied on simulated environments or expensive platforms now have an option that ships with reproducible tooling and an open license. Whether the hardware holds up under sustained training cycles, and whether the community produces shared behaviors the way it has shared language models, will depend on adoption rates that no pre-order number can predict.
What is clear is that Hugging Face is continuing to extend its open-source ethos from software into physical systems. The Microduck is not a research lab prototype. It is a product, priced for purchase, with shipping promised within months. That distinction, more than any single specification, is what makes it notable.
