NVIDIA just built a brain for robots—Groot N1 model explained
What's the story
NVIDIA has unveiled its latest AI foundation model, Groot N1, for humanoid robotics. The announcement was made at the company's GTC 2025 event in San Jose.
This "generalist" model is unique as it has been trained on both synthetic and real data.
According to NVIDIA, Groot N1 employs a "dual system architecture" that mimics human cognitive processes by "thinking fast and slow."
Project evolution
Groot N1: A step forward from Project Groot
The launch of Groot N1 also marks the evolution of NVIDIA's Project Groot, which was introduced at last year's GTC conference.
While Project Groot was focused on industrial applications, the new model expands its scope to humanoid robots of all shapes and sizes.
This expansion also shows NVIDIA's commitment to pushing AI technology beyond industrial use cases.
Dual systems
Groot N1's dual system architecture explained
The dual system architecture of Groot N1 includes a slow thinking system and a fast thinking system.
The slow thinking system allows the robot to perceive and reason about its environment and instructions, before planning appropriate actions.
Meanwhile, the fast thinking system translates this plan into robotic actions, including those involving multi-step object manipulation.
Open-source release
Groot N1: A step toward generalist robotics
Along with making Groot N1 available as an open-source model, NVIDIA is also releasing simulation frameworks and blueprints for generating synthetic training data.
"The age of generalist robotics is here," said Jensen Huang, CEO of NVIDIA, emphasizing the importance of this development in the world of humanoid robotics.