Nvidia is releasing what is called the AI foundation model for humanoid robotics.
The model, called Groot N1, announced at GTC 2025 in San Jose, is a “generalist” model trained on both synthetic and real data. In a video introducing the Groot N1, Nvidia states that it features a “dual system architecture” to “think fast and slow” inspired by human cognitive processes.
Groot N1 is an evolution of Nvidia’s Project Groot, which started at the GTC conference last year. While Project Groot was aimed at industrial use cases, Groot N1 expands its focus on humanoid robots in various form factors.
Groot N1’s slow-thinking system allows the robot to recognize its environment and instructions, recognize the reason, and then, according to Nvidia, plan the appropriate action. For fast thinking systems, we translate the above-mentioned plans into robotic actions. This includes actions that manipulate objects in multiple steps.
Groot N1 is open source and is available. Alongside the models, Nvidia is releasing a simulation framework and blueprint for generating synthetic training data.
“The era of generalist robotics is here,” Nvidia CEO Jensen Huang said in a statement.
Humanoid robots have attracted a lot of publicity in recent years. Companies like X1 and Figure are trying to create general purpose robots that move more or less like humans. The challenges are formidable, but these companies argue that technology has reached a point where mass-produced humanoid robotic systems have reached a realistic short-term goal.
Many disappointments in recent history of robotic history suggest that it is easier to say than achieved.