95% success rate after 5 demonstrations per task… NVIDIA unveils GR00T 2
Original headline작업별 시연 5회 후 95% 성공률…엔비디아 GR00T 2 공개
Not investment advice. Muhabbit scores how much a story may matter, not which way to trade.
NVIDIA unveiled GR00T 2, its "world action model" robot foundation model, claiming a 95% average success rate across 50 tasks when given five demonstrations per task in internal evaluations. The model takes video, commands and robot state as input and jointly predicts future video and robot actions, with training data including first-person human video and handheld capture. NVIDIA presented GR00T 2 alongside Isaac Lab 3.0 at CoRL 2026 in Austin, Texas, shifting focus from locomotion training to precision manipulation such as cable connection and parts assembly. The disclosure matters because it signals NVIDIA extending its AI stack from data-center GPUs into physical robotics, where demonstration efficiency determines deployment cost. Next, watch for independent verification of the 95% figure and adoption by robot makers building on Isaac Lab 3.0.
Muhabbit's AI wrote this summary in its own words from TokenPost's report and translated it from Korean. It can get details wrong, so check the source.
Named in this story
NVIDIA (NVDA), GR00T 2, Isaac Lab 3.0, CoRL 2026, Austin, Texas
A linked name opens that asset's sentiment page.