NVIDIA and Hugging Face Bring New Models and Frameworks to LeRobot for the Open Robotics Community

NVIDIA and Hugging Face are collaborating to integrate NVIDIA Isaac GR00T 1.7, an open VLA (Vision-Language-Action) model for humanoid robots, and NVIDIA Isaac Teleop, a data collection framework, into LeRobot, Hugging Face's open source robotics library.

Through this collaboration, developers will gain a unified pipeline that lets them collect, train, fine-tune, and deploy robot data through standardized workflows. Going forward, NVIDIA Cosmos 3, a frontier model for physical AI, will also be added to LeRobot.

Key integrations include the following:

• NVIDIA Isaac Teleop: A framework that captures high-quality human demonstration data via external devices and shares it in a standardized format
• NVIDIA Isaac GR00T 1.7: A robot foundation model that helps post-train and deploy models for new robot embodiments and tasks
• NVIDIA Cosmos 3 (coming soon): A world foundation model that generates and augments robot data and simulates scenarios in data-scarce situations

In addition, the collaboration supports the entire robot development process through the open source physical AI dataset, which has been downloaded more than 15 million times, simulation frameworks based on NVIDIA Isaac Sim and Isaac Lab, and NVIDIA Jetson Thor integration.