Nvidia has unveiled an AI foundation model designed specifically for humanoid robotics, which it refers to as a significant development in the field.

The announcement was made at the GTC 2025 conference in San Jose, where the company introduced the model, named Groot N1. This model is classified as a “generalist” model, as it has been trained on a combination of synthetic and real-world data. According to a video presentation by Nvidia, Groot N1 features a unique “dual-system architecture” that enables both rapid and slow thinking, mirroring the cognitive processes of humans.

Groot N1 represents the next stage in the evolution of Nvidia’s Project Groot, which was initially launched at the company’s GTC conference the previous year. While Project Groot focused primarily on industrial applications, Groot N1 expands its scope to encompass humanoid robots in various forms and sizes.

The slow-thinking component of Groot N1 allows a robot to perceive its surroundings, understand instructions, and devise a plan of action, as explained by Nvidia. In contrast, the fast-thinking component executes the planned actions, including those that involve manipulating objects over multiple steps.

Nvidia has made Groot N1 available as an open-source model. Additionally, the company is releasing simulation frameworks and guidelines for generating synthetic training data to support the model.

Nvidia CEO Jensen Huang stated, “The era of generalist robotics has begun.”

In recent years, humanoid robots have garnered significant attention. Companies such as X1 and Figure are working towards developing general-purpose robots that mimic human movements. Although the challenges are substantial, these companies claim that current technology has advanced to the point where mass production of humanoid robotic systems is a feasible near-term goal.

However, numerous setbacks in the recent history of robotics, such as disappointing developments, suggest that achieving this goal may be more challenging than anticipated.