GEAR-SONIC

GEAR-SONIC (SONIC) is a humanoid behavior foundation model developed by NVIDIA, hosted on Hugging Face at nvidia/GEAR-SONIC. SONIC stands for Supersizing Motion Tracking. Rather than building separate controllers for predefined motions, SONIC uses motion tracking as a scalable training task, learning from large-scale human motion data to produce a single unified policy for natural whole-body movement. The model supports walking, running, crawling, jumping, manipulation, and more. The release includes real-time VR whole-body teleoperation via a PICO headset, a C++ inference stack for real-time control on humanoid robots, and a kinematic planner for real-time locomotion generation with multiple movement styles. The model card documents SONIC v1.1 checkpoints and is distributed under the NVIDIA Open Model License. A related paper is available at arXiv:2511.07820.