Hugging Face Launches $399 Open-Source Microduck Robot

Hugging Face, in collaboration with Pollen Robotics, has officially unveiled the Microduck, an open-source desktop robot priced at $399 and scheduled for delivery before Christmas. Standing 25 centimeters tall, the lightweight device is engineered to democratize access to physical artificial intelligence and world model development. According to CEO Clem Delange, the Microduck represents a strategic step toward making advanced robotics accessible to researchers and hobbyists alike. The robot is equipped with a forward-facing camera, lidar sensors, and dual inertial measurement units, enabling spatial awareness and dynamic movement. Its mechanical design allows it to waddle, crouch, recover from falls, roller skate, and manipulate objects weighing up to 800 grams using its beak. Rather than relying on predefined behaviors, the Microduck is built for reinforcement learning. Developers can train behaviors in simulation and deploy them directly to the physical unit using a fully open software development kit hosted on GitHub. This sim-to-real pipeline supports continuous fine-tuning and redeployment, aligning with Hugging Face’s broader mission to standardize open-source AI workflows across software and hardware. The product launch extends Hugging Face’s hardware initiatives, which began with the April 2025 acquisition of French robotics startup Pollen. The company’s existing desktop portfolio includes the $499 Reachy Mini and the $399 Reachy Mini Lite, which runs on standard Mac or PC hardware. The Microduck continues this trajectory by offering a compact, low-cost platform for testing embodied AI models. Delange emphasized that open-source architectures provide superior transparency and auditability compared to proprietary black-box systems, a stance particularly relevant given recent concerns over centralized AI control. The rollout occurs amid significant strategic developments for the company. Reports indicate that Hugging Face is in advanced talks to be acquired by Nvidia at a $13 billion valuation, a move that would solidify both firms’ shared commitment to open-source infrastructure and accelerated AI computing. This partnership follows years of collaborative backend support provided by Nvidia since 2023. However, the hardware launch also follows heightened scrutiny regarding platform security, including a recent incident where an OpenAI safety sandbox compromise exposed internal servers. While open-source robotics mitigate proprietary data monopolies, Delange acknowledged that third-party applications built on the platform could still introduce privacy vulnerabilities if not properly audited. With full training stacks and simulation environments available to the public, the Microduck positions Hugging Face at the forefront of accessible physical AI. The device is expected to serve as a testing ground for researchers exploring reinforcement learning, sensor fusion, and decentralized robotics, reinforcing the company’s shift from a model-hosting platform to a comprehensive AI hardware and software ecosystem.