Nvidia has firmly embedded itself in the future of robotics by announcing the availability of the Isaac GR00T N1 foundation model during its GTC 2025 event. This innovative, open-source model is set to accelerate the development and functionality of humanoid robots. “The age of generalist robotics is here,” stated Jensen Huang, Nvidia’s founder and CEO, emphasizing the transformative impact of this technology. The GR00T N1 model, paired with advanced data-generation frameworks, is designed to empower developers, allowing them to push the boundaries of artificial intelligence in robotics.
In a striking demonstration, Huang showcased the capabilities of 1X’s NEO Gamma humanoid robot using the GR00T N1 model, demonstrating its ability to autonomously perform tasks such as tidying up environments through adept manipulation of its surroundings. Bernt Børnich, CEO of 1X Technologies, highlighted the significant leap in robot reasoning and skills afforded by Nvidia’s model. Børnich emphasized how a minimal amount of post-training data enabled the full deployment of the NEO Gamma, reinforcing the importance of making robots that are not mere tools, but rather companions capable of assisting humans in meaningful ways.
The GR00T N1 model, initially introduced as Project GR00T last year, is based on a dual-system architecture that mirrors human cognition. It comprises a fast-thinking action model, termed System 1, which functions like human reflexes, and a slow-thinking model, dubbed System 2, that employs a vision-language model for reasoning and task planning. The seamless collaboration of these systems allows for precise and flexible robot movements, vital for complex tasks that require a combination of basic skills.
Nvidia’s innovative vision expands the potential for humanoid robotics. The company has enabled other organizations like Boston Dynamics and Agility Robotics to access the GR00T N1 model, furthering the development of humanoid robots that promise advancements in various practical applications. Additionally, Nvidia provides easily accessible training data and task evaluation scenarios available for download on platforms such as Hugging Face and GitHub, encouraging developers to tailor the robots to meet specific needs and enhance their operational capacities.
Huang’s keynote illustrated that the technical advancements represented by the GR00T N1 model could usher in significant developments across multiple robotic applications. The convergence of AI, robotics, and advanced human interaction methods places Nvidia at the forefront of innovation in this evolving technology landscape. By amplifying robot capabilities through adaptable models like the GR00T N1, Nvidia is laying the groundwork for a future where humanoid robots become integral to everyday life, transforming how they perform tasks and support human activities across all walks of life.