NVIDIA has released new open-source models and simulation libraries that should accelerate the development and deployment of physical AI.
Revealed at Conference on Robot Learning 2025, the launch includes the Newton Physics Engine, Isaac GR00T N1.6 foundation model, and Cosmos simulation libraries.
“With these latest updates, developers now have the three computers to bring robots from research into everyday life — with Isaac GR00T serving as robot’s brains, Newton simulating their body and NVIDIA Omniverse as their training ground,” said Rev Lebaredian, Vice President of Omniverse and Simulation Technology at NVIDIA.
Co-developed with Google DeepMind and Disney Research, Newton Physics Engine has been integrated into NVIDIA Isaac Lab. It enables high-fidelity simulations of complex robotic behaviours such as humanoid locomotion, dexterous manipulation and multi-agent coordination.
Roboticists can now train and test systems in virtual environments with unprecedented realism so learned skills transfer reliably to physical robots. Enterprises operating in dynamic sectors such as manufacturing, logistics and healthcare, benefit from faster prototyping, safer deployment and reduced operational risk.
Complementing this is the Isaac GR00T N1.6 model, which brings human-like reasoning to robots. Capable of interpreting vague instructions and applying prior knowledge, GR00T lets robots generalise across tasks and environments.
Integrated with the Cosmos Reason vision-language model, GR00T enhances robotic understanding of multi-modal inputs for better adaptability to real-world scenarios. The results are smarter service robots, intuitive home assistants, and more responsive automation in public spaces.
Cosmos World Foundation Models, which has been downloaded more than three million times, further streamline development by enabling scalable data generation from text, image and video prompts.
Developers can simulate entire workflows, generate synthetic training data, and fine-tune models within NVIDIA’s Omniverse ecosystem. The R&D cycle is dramatically shortened while the barrier to entry for robotics innovation are lowered.
NVIDIA’s platform supports scalable infrastructure with GB200 NVL72, RTX PRO Servers and Jetson Thor for on-device inference, enabling enterprises to deploy intelligent robots across edge and cloud environments. The open-source nature of these tools fosters collaboration and standardisation for faster adoption and integration into existing systems.
For end users, this translates into more capable and trustworthy robots in everyday life. Whether assisting in hospitals, navigating warehouses or supporting households, these advancements bring physical AI closer to human-level interaction and reliability.
NVIDIA robotics-related technologies have been adopted by leading research labs and institutions such as Carnegie Mellon, University of Washington, ETH Zurich, and National University of Singapore.
Entelechy Asia 