GR00T 2
$ 0
GR00T 2 is NVIDIA’s open world action model for robots. It learns largely from human video, predicts how scenes will change and works across arms, dexterous hands and humanoids.
Available on backorder
Brain Score
7
Specifications and details:
| Nationality | US |
|---|---|
| Website | https://www.nvidia.com |
| Model type | Foundation Model |
| Manufacturer | NVIDIA |
| Release date | November 9, 2026 |
Description
GR00T 2 is the newest robot foundation model from NVIDIA, built by its GEAR research lab. NVIDIA unveiled it on November 9, 2026, at the Conference on Robot Learning in Austin, Texas. It follows the company’s earlier GR00T N1 models. However, GR00T 2 marks a clear shift in approach. NVIDIA calls it a “world action model.” In simple terms, it predicts how a scene will change before the robot moves, then acts on that prediction.
The biggest change is how it learns. Most of its training comes from first-person videos of people doing everyday tasks. As a result, robots can learn skills from human demonstrations and share them between different bodies. NVIDIA tested it on robot arms, dexterous hands and the Unitree G1 humanoid. For example, robots assembled toy trucks and handled delicate objects with five-fingered hands. A lighter version also runs directly on the humanoid while it walks and carries items. Meanwhile, companies such as Foxconn and Siemens have started testing the model in factories.
New! 2026 Humanoid
Robot Market Report
198 pages of exclusive insight from global robotics experts — uncover funding trends, technology challenges, leading manufacturers, supply chain shifts, and surveys and forecasts on future humanoid applications.
now Google DeepMind
Contact Humanoid.guide
Website: https://research.nvidia.com/labs/gear/gr00t-2/
