NVIDIA Releases New Tools for Physical AI Developers

NVIDIA unveiled open-source physical AI agent skills and tools at GTC Taipei and Computex, aimed at accelerating development for robotics and autonomous vehicles. These tools are expected to reduce the cost, time, and complexity of building physical AI workflows.

Tue Jun 02 2026

NVIDIA Unveils Open-Source Tools for Physical AI

NVIDIA has released open-source physical AI agent skills and tools at GTC Taipei and Computex. These are designed to assist developers in robotics, autonomous vehicles, visual AI, and industrial digital twins. The company states that these tools can help reduce the costs, time, and complexity associated with building large-scale physical AI workflows.

Accelerating Development with AI Agent Toolkit

The new skills, available as part of the NVIDIA Agent Toolkit, will enable AI agents to expedite data generation, simulation, training, evaluation, and deployment pipelines for robots, autonomous vehicles (AVs), factories, and laboratories. Jensen Huang, founder and CEO of NVIDIA, highlighted that "AI agents are revolutionizing software development, and that shift is now coming to physical AI, extending into the systems that will transform transportation, manufacturing, healthcare, and robotics." NVIDIA is optimizing its entire physical AI stack for agents by transforming libraries, models, and frameworks into agent-callable tools. A key component, Cosmos 3, is described as a frontier foundation model for physical AI, capable of understanding videos and text, identifying crucial information, and performing physically accurate simulations and predictions.