NVIDIA Unveils GR00T 2 World-Action Model for Robot Learning
NVIDIA introduced GR00T 2 at CoRL as an open world-action model that learns from human video and robot demonstrations while jointly predicting future observations and actions. The 2-billion-parameter Transformer was trained on 402,730 hours of first-person human video, 10,184 hours of handheld-gripper demonstrations and 23,500 hours of robot teleoperation data. NVIDIA said GR00T 2 completed 40 of 50 manipulation tasks at least once without task-specific demonstrations on the target robot, with average success rates of 71% after one demonstration and 95% after five. The company noted that the tasks were selected because the model already learned them well, limiting the generality of the results. NVIDIA also presented Isaac Lab 3.0 with Newton and said it achieved up to 2.5 times the training throughput of Isaac Lab 2.3 with PhysX in a vision-based dexterous-manipulation benchmark. Code availability was listed for December, while Foxconn, Siemens and Nexuni were testing the model without announcing production deployments.

Nvidia unveiled GR00T 2, an open AI model that helps robots learn delicate hand skills, trained partly on 100,000 hours of human video. It reached 95% average success with five demonstrations per task in Nvidia’s tests. Foxconn, Siemens and Nexuni are testing it.
Nvidia has unveiled GR00T 2, an AI model that helps robots learn delicate hand skills partly by watching videos of people, the company said on Monday. It announced the model at the Conference on Robot Learning in Austin, Texas.
#NVIDIA#GR00T 2#Isaac Lab 3.0#Newton#Human-1M-100K#Cosmos-Predict2.5-2B