Nvidia has just taken a bold step into the future of robotics with the launch of the Isaac GR00T N1. This open-source, customizable model is set to transform how we think about humanoid robots. Announced at GTC 2025, the GR00T N1 is designed to boost both the development and functionality of humanoid robots. During his keynote, Nvidia’s founder and CEO, Jensen Huang, confidently stated, “The age of generalist robotics is here.” This model, alongside cutting-edge data-generation and learning frameworks, is ready to push the boundaries of what robots can achieve.
At the conference, we saw the GR00T N1 in action through 1X’s NEO Gamma robot, which autonomously handled tidying tasks. Bernt Børnich, CEO of 1X Technologies, emphasized the importance of adaptability and learning for the future of humanoids. He pointed out how the GR00T N1 enhances a robot’s reasoning and skills, allowing their NEO Gamma to operate effectively with minimal extra training data. This aligns with their vision of creating robots that are more than just tools—they’re companions that can genuinely assist us in meaningful ways.
Several top robotics companies, including Boston Dynamics and Agility Robotics, have had early access to the GR00T N1. Initially known as Project GR00T, this model uses a dual-system architecture that mirrors human cognitive processes. System 1 serves as a “fast-thinking action model,” similar to human reflexes, trained using human demonstrations and synthetic data from Nvidia’s Omniverse platform. On the other hand, System 2, powered by a vision language model, is a “slow-thinking model” that understands its environment and plans actions, which System 1 then executes with precision.
Developers have the flexibility to further customize the GR00T N1’s behavior for specific tasks by post-training with human demonstration data or simulations. Nvidia has made the GR00T N1’s training data and task evaluation scenarios available on platforms like Hugging Face and GitHub, making it easier for developers to get started.
