$NVDA Unveils GR00T 2 and Isaac Lab 3.0 to Advance AI Robotics

Nvidia ($NVDA) introduced GR00T 2 and Isaac Lab 3.0 at CoRL, unveiling new open robotics models, simulation tools and benchmarks designed to help robots perform more precise physical tasks, including cable handling and small-component assembly.
GR00T 2 is an open-world action model built on Nvidia Cosmos and trained on robot data alongside 100,000 hours of first-person human video. Across 50 evaluated tasks, Nvidia reported that it:
- Completed 40 tasks at least once without task-specific demonstrations.
- Achieved a 71% average success rate with one demonstration per task, increasing to 95% with five demonstrations.
- Transferred learned skills between different robot bodies without robot-specific demonstrations.
Foxconn is applying GR00T 2 to automate server manufacturing, while Siemens is evaluating the model for industrial applications.
Isaac Lab 3.0 introduces an updated robot-learning framework that allows AI agents to help configure training environments and uses the Newton physics engine to simulate flexible objects such as cables and cloth. Nvidia reported up to 2.5x higher training throughput than Isaac Lab 2.3 with PhysX in a dexterous manipulation benchmark.
Nvidia also introduced DexBench and VABAR, two open benchmarks designed to evaluate robotic dexterity, adaptability and performance across real-world industrial tasks.