GR00T-H

Healthcare Physical AI VLA

Commercial

GR00T-H is a post-trained variant of NVIDIA Isaac GR00T N1.7 for surgical robots, adapted with 601 hours of real-world surgical data from the Open-H-Embodiment dataset across 7 robot platforms, including CMR Versius, dVRK, and KUKA LBR iiwa. Cosmos-Reason2-2B encodes camera frames and a text instruction; a flow-matching diffusion transformer combines them with robot proprioception to output continuous motor actions for each embodiment. The 3B-parameter model is intended for robotics R&D such as surgical and robotic-ultrasound policy research and benchmarking, not clinical deployment.

Multi HuggingFace GitHub Paper
No agent skill for this model yet. Use it directly via HuggingFace or GitHub.

Engineering & research use only — not a medical device, not for diagnosis or clinical decision-making. Skills follow the open Agent Skills spec.