NVIDIA unveils GR00T 2 open-world robot action model, hitting 95% success with 5 demos
CyberRobooo · x · 2026-11-09
At CoRL in Austin, NVIDIA unveiled GR00T 2, an open-world action model built on Cosmos and trained on robot and human data, including 100,000 hours of egocentric human video.
Key numbers
- Completed 40 of 50 tasks at least once without task-specific demos; averaged 71% success with one demo and 95% with five.
- Transfers skills across robot bodies; partners are testing laundry handling and Ethernet cable swaps with a YAM arm.
- Simulation is doing more heavy lifting: Dyna uses Isaac Lab to train whole-body control for its semi-humanoid Taku.
UC Berkeley professor Ken Goldberg called agentic robotics "the most exciting technological development since the internet in 1995."
More from Embodied
- Apple's HALO: humanoid trained purely in simulation hits 91% success in real homes — heyshrutimishra · 2026-11-10
- Apple's HALO humanoid trains in 1000 simulated homes, hits 91% success in real houses — heyshrutimishra · 2026-11-10
- RoboMoCap: 200-hour egocentric vision × optical mocap dataset for humanoid robot learning — DJiafei · 2026-11-09
- Boston Dynamics puts wheels on Atlas to stress-test its whole-body RL controller — CyberRobooo · 2026-11-09
- Guava distills GPT-5.4 into a 4B robot agent with just 2,000 simulated trajectories — jbhuang0604 · 2026-11-09
- Dex-One2Many learns dexterous manipulation from one human video, 95%+ success across 4 hands — jbhuang0604 · 2026-11-09