Berkeley's Open-Source Humanoid Robot Costs Less Than $5,000

A $5,000 Humanoid Robot You Can Print at Home

UC Berkeley researchers have open-sourced a complete humanoid robot that anyone can build with a desktop 3D printer and off-the-shelf electronics. The project, called Berkeley Humanoid Lite, costs less than $5,000 in parts. All CAD files, firmware, and training code are available on GitHub.

The robot stands about knee-high and walks on two legs. Its most distinctive feature is a set of 3D-printed cycloidal actuators. These gear-like mechanisms convert motor rotation into smooth, precise movement with very little backlash. The team printed the actuators in PLA, a material most makers have in their filament drawers. That choice keeps costs low and eliminates the need for special industrial equipment.

Built for Classrooms and Research Labs

Berkeley's design philosophy prioritizes accessibility. The largest printable part fits into a 200 mm cube build volume, well within the capacity of any modern desktop printer. Components were selected for availability on Digi-Key and similar sites. No custom machining. No hard-to-find bearings.

The platform includes a full locomotion stack based on reinforcement learning. Engineers trained the robot in simulation and transferred the policy to hardware without retuning. That zero-shot capability makes the platform useful for university courses and research projects where time and budget are limited.

Why It Has Not Gone Viral Yet

Despite the low cost and open-source license, Berkeley Humanoid Lite has not spawned a wave of home builds. A search for build videos on YouTube returns only a handful of results. The research team published its paper in April 2025 and maintains a Discord server and a WeChat group, but community activity remains modest.

Part of the reason is practical. PLA wears quickly on gear surfaces. The researchers acknowledge that the plastic may be the design's weak point over time. Better materials exist, but they either require higher-temperature printers or are not safe to print at home. The team believes that improvements in desktop hardware will close that gap within a few years.

Another factor is funding. Humanoid robotics competes for attention and capital against AI software projects that promise faster returns. Building an open-source physical platform is harder to monetize than releasing a machine learning framework. The researchers behind Berkeley Humanoid Lite are betting that community adoption will grow as more educators and hobbyists discover the project.