1.jpg)
At CES 2026, NVIDIA announced a significant expansion of its physical AI ecosystem, releasing new open models and frameworks while showcasing a new wave of AI-driven robots from global partners. The announcements include the NVIDIA Cosmos and GR00T open models for robot learning, the Isaac Lab-Arena evaluation framework, the OSMO compute orchestration tool, and the new Jetson T4000 module, all designed to accelerate the development of generalist-specialist robots across industries.
NVIDIA releases new open physical AI models: Cosmos (for simulation/reasoning) and GR00T N1.6 (for humanoid robot control).
Global partners Boston Dynamics, Caterpillar, Franka, LG Electronics, and NEURA Robotics unveil next-gen robots built on NVIDIA tech.
New Isaac Lab-Arena framework standardizes large-scale robot evaluation and benchmarking in simulation.
OSMO is a new cloud-native orchestration framework to unify and simplify robot development workflows.
The Jetson T4000 module, based on Blackwell, offers 4x greater energy efficiency for edge AI and robotics.
NVIDIA collaborates with Hugging Face to integrate its models into the LeRobot open-source framework.
NVIDIA positions these advancements as a pivotal moment for physical AI, where models can understand, reason, and plan actions in the real world. By providing a comprehensive stack—from processors and simulation to open models—NVIDIA aims to democratize the development of intelligent machines, enabling partners to transform industries with robots that can learn multiple tasks rather than being programmed for single functions.
Jensen Huang, founder and CEO of NVIDIA, framed the announcements as a breakthrough: “The ChatGPT moment for robotics is here. Breakthroughs in physical AI — models that understand the real world, reason and plan actions — are unlocking entirely new applications. NVIDIA’s full stack of Jetson robotics processors, CUDA, Omniverse and open physical AI models empowers our global ecosystem of partners to transform industries with AI-driven robotics.”
To reduce the massive resource requirements for building robot foundation models, NVIDIA is releasing a suite of open, customizable models on Hugging Face. These include the Cosmos family for world modeling and reasoning, and the GR00T N1.6 model for humanoid robot control. Complementing these are new open-source frameworks: Isaac Lab-Arena for standardized simulation-based evaluation and benchmarking, and OSMO, a cloud-native orchestration tool that simplifies managing complex training and testing pipelines across different compute environments.
The ecosystem showcase is broad, with partners deploying NVIDIA technologies for diverse applications. Boston Dynamics, Humanoid, and NEURA Robotics are integrating the powerful Jetson Thor computer into their humanoids. LG Electronics unveiled a new home robot, and Caterpillar is expanding its AI and autonomy collaboration with NVIDIA. For broader edge robotics, the new Jetson T4000 module brings the Blackwell architecture’s efficiency gains. Furthermore, the collaboration with Hugging Face integrates NVIDIA’s models into the popular LeRobot framework, uniting millions of developers in the open-source physical AI community.
About NVIDIA
NVIDIA is the world leader in AI and accelerated computing.
TechIntelPro empowers B2B technology decision-makers with the latest industry trends, deep insights, leadership perspectives and actionable intelligence to drive both organizational and professional milestones—and thrive on the tech pulse.