Berkeley Humanoid Lite stands at about 1 meter and weighs approximately 32 pounds. UC Berkeley
Berkeley engineers have unveiled the Berkeley Humanoid Lite, a low-cost, fully open-source humanoid robot designed to help beginners enter the world of robotics.
Built from 3D-printed parts and off-the-shelf components, the robot costs under $5,000 and requires no specialized equipment to assemble.
The humamoid is one meter tall and weighs 35.2 pounds (16 kilograms) and features modular actuators with durable cycloidal gears.
According to the team, with full access to hardware designs, code, and training tools, the project empowers hobbyists, students, and educators to customize, build, and improve humanoid systems from the ground up.
Humanoid robotics has advanced significantly, expanding applications into automation, healthcare, research, and personal assistance. Yet, accessibility remains a challenge, especially for newcomers. While commercial robots exist, they are often costly and built with proprietary hardware and software, making repairs and customization difficult.
For enthusiasts and aspiring roboticists, this restricts their ability to experiment and learn. Furthermore, many research projects presume access to sophisticated equipment that most people cannot afford, such as CNC machines and bespoke PCBs.
The Berkeley Humanoid Lite team recognized these obstacles and set out to develop an open-source, reasonably priced, and adaptable robot platform. The project provides an accessible introduction to humanoid robots by utilizing off-the-shelf materials and 3D-printed components, allowing users to construct, alter, and learn without needing specialist tools or financial limitations.
The humanoid is built around a modular 3D-printed gearbox for the actuators and the robot’s body. All additional parts can be easily sourced from common e-commerce sites or made using standard desktop 3D printers. This approach keeps the total hardware cost below $5,000, significantly cheaper than comparable commercial humanoid robots.
According to researchers, if any part breaks or wears out, it can be quickly reprinted and replaced. When assembled, the robot stands about 1 meter tall and weighs around 35.2 pounds (16 kilograms), making it practical and accessible for hands-on development.
Democratizing humanoid robotics
Building the Humanoid Lite takes roughly a week for a beginner, though the amount of time needed can vary depending on experience. Feedback from the community on sites like Discord demonstrates active participation, with members successfully putting their robots together and displaying them.
The team claims that since development started four years ago, the number of reasonably priced metal actuators on the robotics market has increased. However, Berkeley Humanoid Lite’s approachable and modular architecture keeps it ahead of the competition. Users can begin small and work up to entire limbs by building a single actuator and evaluating its motion.
The team designed a cycloidal gearbox for the actuators to compensate for the lower strength of 3D-printed plastics compared to metal. Large gear teeth in this design reduce wear and increase longevity by distributing load across a larger surface area. Numerous tests revealed that none of the actuators have failed in experimental settings, and they function similarly to commercial devices and can endure repeated use. If necessary, a new gear unit can be printed to replace any damaged actuator swiftly.
Technically, the robot can walk and grip objects because of a locomotion controller that uses reinforcement learning and teleoperation based on joysticks. Although walking is still a work in progress, the open-source hardware, firmware, and training resources encourage users to enhance functionality and boost performance, encouraging cooperative growth in humanoid robotics.
“I believe in the spirit of open-source communities, an ecosystem where people share ideas and knowledge. We hope that Berkeley Humanoid Lite will help move us closer to democratizing the development of humanoid robotics,” said Yufeng Chi , a Ph.D. student in the Department of Electrical Engineering and Computer Sciences at UC Berkeley, in a statement.
The details of the team’s research were presented at the 2025 Robotics Science and Systems conference.
MasterCard