Pollen Robotics (part of Hugging Face)

Bordeaux, 2016: that is where the story of Pollen Robotics begins, the French company behind the open-source humanoid platform Reachy, now in its second generation as Reachy 2. From the start the company bet on openness: the robot was meant as a tool for researchers who want to program it, change it and teach it, not a closed product for a single task. In time Pollen was acquired by Hugging Face, which placed Reachy 2 in the LeRobot ecosystem, an open toolkit for robot learning. The acquisition has practical meaning. For teams working on models that learn from demonstrations it means a ready path from hardware to software, and for the company itself access to a community that already has deep experience with open models. The manufacturer gives no data on funding or headcount, so we do not describe the scale of the enterprise. What is known is that it works at the meeting point of robotics and artificial intelligence, the stream known as embodied AI, which we discuss in “VLA Models: How AI Controls Humanoid Robots in 2026”. The general principles of building such machines are explained in “How Does a Humanoid Robot Work”, and the technical terms are collected in “Robotics Glossary: 65 Terms with Polish Equivalents”. This story is worth reading as an example of the European approach to research robotics.

The company's showpiece is the Reachy 2, a mobile research torso 151 cm tall and weighing 50 kg. It has arms with seven degrees of freedom (DOF), and the manufacturer gives the payload as 3 kg. How to understand such parameters is covered in “Degrees of Freedom (DOF) in Humanoid Robots”. The robot runs ROS 2 and works with LeRobot, which eases work on manipulation and human-robot interaction. The price according to April 2025 data is USD 70,000, about PLN 271,300 (≈ EUR 62,170 at NBP rate 4.3641 of 21 October 2026), and orders are placed by email. This figure should be treated as indicative, since it excludes tax and shipping costs. The Reachy 2 is intended for research into embodied AI, manipulation and HRI, so its natural places of work are a laboratory and a university. It is not a robot for a warehouse hall or shift work. The whole class of machines is covered by the category “Humanoid robots”. Before buying it is worth working out whether the team has the skills to use an open platform to the full.

Unlike Chinese and American rivals that focus on industrial deployments, Pollen addresses mainly researchers and builders of embodied-AI solutions. It sits closest to Hugging Face, its own ecosystem, in which the cheaper SO-101 operates. Among research platforms it also competes with PAL Robotics, represented by the PAL Robotics ARI, and with Trossen Robotics. What distinguishes it is openness: whoever buys a Reachy 2 gets access to tools that let them adapt the robot to their own research. Its weak side is the price, high compared with cheaper arm kits, and the need for in-house integration skills. The manufacturer gives no sales volume, so we cannot judge how widely the Reachy 2 is used. The company's position is, however, strengthened by belonging to Hugging Face, and the platform's open nature favors experiments that closed commercial designs do not allow.

The manufacturer gives no information on sales or distribution of the Reachy 2 in Poland, and no Polish deployments are documented. Orders are placed directly by email, which means a Polish university or laboratory buys without a local intermediary. For domestic research teams the Reachy 2 may be an interesting alternative to closed commercial platforms, although it requires in-house integration skills. Helpful guidance is in “How to Start Learning Robotics in Poland”. On import, tax and transport costs have to be added. Universities can check whether the purchase can be financed from a research grant, which we cover in the funding section of the site.

The Reachy 2 is a good choice for universities and laboratories that want an open platform for research on manipulation and embodied AI and have a budget of around USD 70,000. It is not suited to production work. Compare it with PAL Robotics and the cheaper SO-101. The biggest plus is openness and the link to the LeRobot ecosystem, the biggest minus the high price. If your budget is smaller, start with cheaper kits and only then consider the full platform.