NVIDIA and Hugging Face have joined forces to enrich LeRobot, an open-source robotics library, by adding NVIDIA’s Isaac GR00T 1.7 and Isaac TeleOp frameworks. These additions provide robotics developers with streamlined tools to collect data, train, validate, and deploy AI models tailored for humanoid and physical robot applications.

Isaac GR00T 1.7, a vision-language-action model designed specifically for humanoid robots, serves as the first commercially viable robot foundation model available on LeRobot. It enables developers to fine-tune and adapt AI policies for new robots and tasks using standardized workflows. Alongside this, Isaac TeleOp offers an open-source framework that simplifies the process of capturing human demonstrations and translating them into high-quality, shareable datasets.

This collaboration aims to bridge NVIDIA’s robotics expertise, encompassing three million developers, with Hugging Face’s extensive AI community of sixteen million users. By integrating these frameworks into LeRobot’s environment, developers gain access to comprehensive tools for data standardization, model training, simulation-based validation, and deployment—all within an open and collaborative ecosystem.

Looking ahead, NVIDIA plans to introduce Cosmos 3, a frontier world foundation model for physical AI, to the LeRobot platform. This model will augment developers’ ability to generate synthetic robotics data, run complex scenario simulations, and further advance AI-driven control systems for robots.

LeRobot functions as a central hub for training, sharing, and running robotics datasets, models, and policies with open workflows. The integration of NVIDIA’s toolsets enhances its mission by reducing resource fragmentation and lowering entry barriers for robotics innovation. Developers can now more easily build on shared models and data, accelerating research translation into practical robot applications.

  • Isaac TeleOp enables standardized robot data collection from external human demonstrations.
  • Isaac GR00T 1.7 provides a fine-tunable foundation model optimized for humanoid robot tasks.
  • Upcoming Cosmos 3 model will support data augmentation and scenario simulation in physical AI.