Hugging Face Launches $2,500 DIY Humanoid Robot

As a general rule, technology is becoming more accessible as time goes by. Computers, smartphones, and high-speed internet, for instance, were once considered luxury items. Today, billions of people around the globe have access to these technologies. Now, if only robotics would follow suit. Sure, hobbyist-grade line-following robots and the like are very inexpensive, but a capable humanoid robot still often costs more than a car.

The LeRobot Humanoid is an open, low-cost, 3D-printed humanoid robot that was designed to make this technology more accessible. Technically, the LeRobot Humanoid is just a pair of legs, but we have to start somewhere, right? And “low-cost” is a relative term — in this world, its $2,500 price tag is a bargain, but that is still a big enough hurdle to keep most people from joining in on the fun. However, if you do have some spare cash and a 3D printer, the LeRobot Humanoid may be the best learning and experimentation platform out there.

Developed by Hugging Face, the project is targeted at researchers, students, and hobbyists who want a humanoid robot that they can build, repair, modify, and understand. It is not quite a polished consumer product, but the team has released what it describes as a full-stack DIY robotics platform. That includes not only the hardware design itself, but also assembly instructions, wiring diagrams, simulation tools, calibration utilities, identification pipelines, and machine learning training environments.

The robot is constructed primarily from 3D-printed parts combined with off-the-shelf electronics and affordable actuators. This approach was taken to maximize reproducibility and experimentation potential. If a structural component breaks, it can simply be reprinted. If a design decision proves problematic, builders can quickly revise and test a new version without waiting for expensive custom parts.

The current release focuses on a bipedal lower-body platform capable of standing and locomotion. While it will not be sprinting through obstacle courses anytime soon, the platform was designed to help researchers close the gap between simulation and real-world robotics. A major challenge in robotics development is that behaviors trained in simulation often fail when transferred to physical hardware. LeRobot Humanoid attempts to address this by connecting design tools, simulation environments, real-world data collection, and parameter identification systems into one continuous workflow.

The project is split across several repositories, each targeting a specific part of the development pipeline. Separate repositories handle mechanical and control co-design, physical hardware documentation, robot model assets, runtime deployment, and simulation-based identification. Training environments are also included through the “lerobot-legged-zoo,” allowing users to develop and evaluate locomotion policies before risking damage to the real hardware.

For now, LeRobot Humanoid remains an experimental research platform rather than a consumer-ready robot companion. But by making humanoid robotics cheaper, more open, and easier to reproduce, projects like this could help bring advanced robotics experimentation to a much wider audience.