NVIDIA unveiled a comprehensive portfolio of new technologies designed to accelerate the development of humanoid robots, led by NVIDIA Isaac GR00T N1-the world’s first open, fully customizable foundation model built for generalized humanoid reasoning and skill execution.
The announcement also introduces advanced simulation frameworks and development blueprints, including the NVIDIA Isaac GR00T Blueprint for synthetic data generation and Newton, a new open-source physics engine purpose-built for robotics development. Newton is being developed in collaboration with Google DeepMind and Disney Research to provide high-fidelity physical simulation tailored to robotic learning and interaction.
Available now, GR00T N1 is the first release in a planned family of pretrained, customizable models that NVIDIA will make available to robotics developers worldwide. These models are designed to help address widespread labor challenges across industries, where global workforce shortages are estimated to exceed 50 million people.
“The age of generalist robotics is here,” said Jensen Huang, founder and CEO of NVIDIA. “With NVIDIA Isaac GR00T N1 and new data-generation and robot-learning frameworks, robotics developers everywhere will open the next frontier in the age of AI.”
GR00T N1 Advances the Humanoid Robotics Ecosystem
The GR00T N1 foundation model introduces a dual-system architecture inspired by human cognitive processes. “System 1” functions as a fast-response action model, similar to human reflexes or intuition, while “System 2” provides slower, more deliberate reasoning for planning and decision-making.
Also Read: Google Cloud Enhances AlloyDB AI for Near-Perfect Text-to-SQL Accuracy
System 2 leverages a vision-language model to interpret its surroundings and understand instructions, enabling it to plan appropriate actions. System 1 then converts these plans into precise, continuous movements. Training for System 1 combines human demonstration data with large volumes of synthetic data generated using the NVIDIA Omniverse™ platform.
GR00T N1 can handle many everyday tasks. It excels at object grasping, single and dual-arm manipulation, and hand-to-hand object transfers. It can also handle multistep tasks. These tasks need long-context reasoning and coordination of different skills. These features assist with real-world tasks. They include material handling, packaging, inspection, and other industrial and service applications.
Robotics developers and researchers can post-train GR00T N1. They can use real-world or synthetic data. This helps customize the model for specific humanoid robots, environments, or tasks.
During his GTC keynote, Huang showcased a live demonstration of 1X’s humanoid robot autonomously completing household tidying tasks using a post-trained policy built on GR00T N1. The demonstration highlighted the results of a collaborative AI training effort between 1X and NVIDIA.
“The future of humanoids is about adaptability and learning,” said Bernt Børnich, CEO of 1X Technologies. “While we develop our own models, NVIDIA’s GR00T N1 provides a significant boost to robot reasoning and skills. With minimal post-training data, we fully deployed on NEO Gamma — advancing our mission of creating robots that are not just tools, but companions capable of assisting humans in meaningful, immeasurable ways.”