Welcome, GR00T 2!

GR00T 2 is a robot foundation model from NVIDIA, developed by the company’s GEAR research lab. It is software for robotics rather than a physical humanoid, robot arm or robotic hand. NVIDIA describes its model category as a “world action model,” distinguishing its positioning within the broader field of robot intelligence. GR00T 2 follows the earlier GR00T N1 models and represents a new direction in NVIDIA’s robotics research portfolio. Its release date is November 9, 2026, with an unveiling at the Conference on Robot Learning in Austin, Texas. It carries a Humanoid.Guide Brain Score of 7 out of 10, placing it in the capable or mature range of that rating system. That overall score is a summary assessment, not a specification for any particular robot body. The distinction matters because a foundation model belongs to the software layer of a robotics system, while hardware, integration and deployment requirements remain separate considerations for organizations evaluating a complete solution.

For technical evaluation, GR00T 2 belongs in the robot intelligence category rather than among mechanical components. A foundation model and a complete robotic system are different products, and that distinction should guide comparisons, procurement and deployment planning. Its overall Brain Score provides a concise assessment, but it should not be interpreted as a measurement of mechanical dexterity, lifting strength or movement speed. Those are system characteristics that require their own evidence and evaluation. Teams considering robotics software should define the intended application, identify the hardware involved and establish clear acceptance criteria before making deployment decisions. This keeps the assessment centered on the complete application rather than treating a model rating as a universal performance guarantee.

GR00T 2 is listed as available on backorder, with an interest option for prospective users. The displayed price is zero dollars, but that listing alone does not establish licensing terms, commercial usage rights or the total cost of deployment. Foxconn and Siemens have started testing the model in factories, connecting its development with industrial evaluation. Such testing is distinct from a confirmed production rollout or a guarantee of commercial readiness for every application. Its broader significance lies in NVIDIA’s continued investment in robot foundation models through GEAR and the progression beyond the GR00T N1 family. For prospective adopters, availability terms and the scope of industrial testing remain practical considerations alongside the model’s overall rating.