What Is a Humanoid Foundation Model? An Introduction to GR00T N1
summary
NVIDIA’s GR00T N1 is an open Vision‑Language‑Action foundation model that trains end‑to‑end on human videos, robot trajectories and synthetic sims and runs on a full‑size humanoid doing bimanual manipulation. Architecture pairs a slow symbolic System 2 planner with a fast motor System 1 at ~120 Hz, uses diffusion transformer blocks plus a VLM encoder and an embodiment‑specific action decoder, and ships as a 2 billion parameter base that adapts via imitation and reinforcement learning across an OVX / DGX / AGX three‑computer lifecycle.
AI-written summary · about
chapters (5)
related talks
16mKeynote: Robot Foundation Models
19mThe Physical Turing Test: Solving General Purpose Robotics
17mBeyond Automation: AI and the Evolution of Robotic Intelligence
01mhumanoid robots and the future of physical AI
15mLeRobot – Lowering the entry barrier to AI for robotics
21mFuture in Motion: Robotics, Self-Driving Cars, and Humanoids Shaping Tomorrow
32mVideo Action Models
27mFrom perception to action: World Foundation Models (WFMs) for autonomous robots and vehicles