Groot N1: Nvidia has made an exciting announcement that its open-source, pre-trained AI foundation model designed for humanoid robots is now publicly accessible. This cutting-edge model, dubbed the Nvidia Isaac GR00T N1 (or Groot N1 for short), was unveiled during Nvidia’s prominent annual GTC conference held in San Jose, California, on March 18.
During the keynote address at GTC 2025, Nvidia CEO Jensen Huang declared, “The age of generalist robotics is here.” He emphasized that with the introduction of Isaac GR00T N1or Groot N1 and innovative data-generation and robot-learning frameworks, robotics developers are poised to explore the frontier of AI like never before.
The foundational model has been rigorously trained on a combination of existing datasets alongside synthetically generated data. In a move to foster innovation, Nvidia will release simulation frameworks and blueprints to facilitate the generation of synthetic training data, further empowering developers in the field.
As companies like X1 and Figure work diligently to refine their humanoid prototypes, major players such as Nvidia and Google are stepping up to provide the essential AI technologies that will drive these advancements. Notably, Google DeepMind recently launched Gemini Robotics, a suite of AI models engineered to equip robots with heightened accuracy and dexterity for complex tasks.
According to a video shared by Nvidia on its YouTube channel, the Groot N1 features a “dual-system architecture” inspired by human cognitive processes, facilitating “thinking fast and slow.” Nvidia asserts that this slow-thinking system will endow humanoid robots with enhanced perception and reasoning abilities, allowing them to make informed decisions on the appropriate actions to undertake. Conversely, the fast-thinking system translates these strategic actions into fluid robotic movements, enabling complex manipulations over multiple steps.
With the introduction of Nvidia’s Isaac GR00T N1 or Groot N1, the future of humanoid robotics is not just on the horizon; it’s now within reach. This groundbreaking advancement is set to transform how robots interact with the world, opening up new possibilities in AI and robotics for years to come.
Unlock the Future
Exciting advancements are here! A groundbreaking foundation model is now available, designed to bring fully customizable, generalized skills and reasoning capabilities to humanoid robots. This innovative technology is set to transform the robotics landscape, enhancing how robots interact with their environments and perform complex tasks.
In a remarkable collaboration, industry leaders NVIDIA, Google DeepMind, and Disney Research have joined forces to develop the next-generation open-source Newton Physics Engine. This powerful engine will propel the next wave of robotics by providing the necessary tools for simulating realistic environments and interactions.
Additionally, a new Omniverse Blueprint for synthetic data generation has been introduced, aiming to jumpstart a robust open-source dataset. This initiative is designed to create a dynamic physical AI data flywheel, fostering rapid advancements in AI and robotics. By leveraging synthetic data, developers can accelerate their projects and unlock the full potential of intelligent automation.
With these cutting-edge resources at their fingertips, robotics developers are poised to create advanced humanoid robots capable of impressive reasoning and skillfulness. The future of robotics is bright, and these powerful tools will pave the way for unprecedented innovations in the field.
A New Era in Humanoid Robotics with Cutting-Edge Technologies
NVIDIA has officially launched a suite of revolutionary technologies aimed at propelling humanoid robot development to new heights. Central to this initiative is the NVIDIA Isaac GR00T N1 or Groot N1, hailed as the world’s first open, fully customizable foundation model designed to enhance generalized reasoning and skills in humanoid robots. This innovative platform also introduces powerful simulation frameworks and tools like the NVIDIA Isaac GR00T Blueprint for synthetic data generation, along with Newton, an open-source physics engine created in collaboration with Google DeepMind and Disney Research.
A Transformation in Robotics Development
Available now, GR00T N1 marks the beginning of a family of customizable models that NVIDIA will pre-train and release to robotics developers globally. This launch is particularly vital in addressing the widespread labor shortages, currently affecting over 50 million individuals worldwide. “The age of generalist robotics is here,” stated Jensen Huang, founder and CEO of NVIDIA. “With NVIDIA Isaac GR00T N1 or Groot N1 and our advanced data-generation frameworks, developers will enter a new frontier in AI.”
Innovative Dual-System Architecture
The GR00T N1 model features a unique dual-system architecture, inspired by human cognitive processes. It consists of two key components:
- System 1: A fast-thinking action model that mimics human reflexes and intuition.
- System 2: A deliberately structured model that supports methodical decision-making.
This setup allows GR00T N1 or Groot N1 to process environmental information and translate plans into precise, continuous movements. The model is trained using a vast array of human demonstration data combined with synthetic data derived from the NVIDIA Omniverse™ platform. GR00T N1 or Groot N1seamlessly generalizes across common tasks such as grasping and transferring objects, making it ideal for various applications like material handling, packaging, and inspection.
Developers can further refine GR00T N1 by post-training it with real or synthetic data tailored to their specific humanoid robots and tasks. During the GTC keynote, Huang showcased a humanoid robot from 1X Technologies successfully performing domestic tidying tasks, thanks to a post-trained policy utilizing GR00T N1.
Leading the Charge in Robotics Collaboration
The partnership between NVIDIA, Google DeepMind, and Disney Research is set to enhance robotic capabilities even further. Together, they are developing Newton, an open-source physics engine engineered to help robots tackle complex tasks with exceptional precision. Built on the advanced NVIDIA Warp framework, Newton will be compatible with renowned simulation platforms like Google DeepMind’s MuJoCo and NVIDIA Isaac™ Lab.
In a significant move to advance robotics research, Google DeepMind and NVIDIA are also creating MuJoCo-Warp, which is expected to accelerate machine learning workloads by over 70 times. Disney Research plans to utilize Newton to enhance its robotic character platform, bringing to life next-generation entertainment robots, including the captivating Star Wars-inspired BDX droids.
Kyle Laughlin, senior vice president at Walt Disney Imagineering Research & Development, expressed enthusiasm about the collaboration: “We’re committed to creating more expressive robotic characters that engage our guests in unprecedented ways.”
Advancing Humanoid Robotics with Synthetic Data
One of the challenges in robot development lies in accessing large, high-quality datasets. To overcome this, NVIDIA has introduced the Isaac GR00T Blueprint for synthetic manipulation motion generation. This revolutionary tool enables developers to generate vast amounts of synthetic motion data from minimal human demonstrations, effectively addressing data limitations in humanoid training.
To illustrate its capabilities, NVIDIA generated 780,000 synthetic trajectories—equivalent to roughly 6,500 hours of human demonstration data—in just 11 hours. By combining this synthetic data with real-world data, GR00T N1 saw a remarkable performance boost of 40%.
Get Started with NVIDIA Technologies Today
NVIDIA is releasing GR00T N1 training data and task evaluation scenarios to the global developer community via Hugging Face and GitHub. The Isaac GR00T Blueprint for synthetic manipulation motion generation is also available as an interactive demo or for download.
To further support developers, NVIDIA introduced the DGX Spark personal AI supercomputer, a turnkey solution designed to accelerate GR00T N1’s capabilities for diverse tasks and environments with minimal programming effort. The Newton physics engine is anticipated to roll out later this year.
For those eager to dive deeper into these groundbreaking technologies, the NVIDIA GTC keynote and the upcoming Humanoid Developer Day sessions will provide invaluable insights. Sessions will include topics like “An Introduction to Building Humanoid Robots,” “Insights Into Disney’s Robotic Character Platform,” and “Announcing Mujoco-Warp and Newton,” showcasing how AI models can train humanoids for real-world applications.
As we stand on the cusp of a new era in robotics, NVIDIA’s innovations like the Isaac GR00T N1 or Groot N1 promise to reshape industries and redefine the capabilities of humanoid robots.