At GTC 2025, Nvidia unveiled Groot N1, an AI foundational model specifically designed for humanoid robots.

Groot N1 is characterized as a "general-purpose" model, trained on both synthetic and real-world data. It employs a "dual-system architecture," inspired by human cognitive processes, to enable a "fast and slow" thinking approach.

Compared to Project Groot introduced at the previous GTC conference, Groot N1 has expanded its application scope beyond industrial uses to support various types of humanoid robots.

The slow-thinking system in Groot N1 focuses on perceiving and interpreting environments and commands, planning corresponding actions. Meanwhile, the fast-thinking system translates these plans into specific robot movements, including multi-step object manipulation.

Nvidia has open-sourced Groot N1 and simultaneously released a simulation framework along with blueprints for generating synthetic training data.

In recent years, humanoid robots have gained significant attention. Companies like X1 and Figure are actively developing versatile robots capable of mimicking human movements. Despite numerous challenges, these companies believe that technology has advanced enough to enable mass production of humanoid robots in the near future.

However, given the many setbacks in the history of robotics development, achieving this goal may not be straightforward.