γ₀: A Generalist Policy for Multi-Embodiment Motion Control

Our collection contains 223 realistic robot models. We train shared reinforcement learning policies across billions of randomized embodiments.
Our vision is a foundation motion policy that improves as our robot collection grows and transfers to new hardware with no robot-specific tuning.
Contribute your robot and help us build the largest open-source collection for multi-embodiment training and deployment. More info below.

Open the interactive playground

Real-world deployments

The same policy controls all four robots.

Unitree Go2 · Silver Badger · Unitree G1

Unitree H1

Contribute to γ₀

Upload and test your own robot directly in the Playground, contribute it to the open-source collection, or work with us on training and real-world deployment.

Contributions of new robot models are cited or acknowledged, and contributions to training and real-world deployment of new robots may qualify for authorship.

Submit your robot model or contact us about real-world deployment by October 15 at contribute@gamma-zero.com.

We review proposals individually for model quality, research fit, safety, hardware access, and our available capacity.

01

Contribute your robot model

Provide a valid URDF of your real robot with proper meshes, inertial parameters, joint limits, actuator torque and velocity limits. Accepted models become part of the open-source collection. Model authors will be cited or acknowledged.

02

Train and deploy with us

Necessary contributions to the training pipeline and the final real-world deployment on a robot, which we couldn't deploy on before, may qualify for authorship.