NVIDIA and Hugging Face Bring Isaac GR00T 1.7 and Teleop to LeRobot
The July 7 collaboration pushes NVIDIA Isaac GR00T 1.7, Isaac Teleop and, soon, Cosmos 3 into Hugging Face's open LeRobot library, giving 3 million robotics developers and 16 million AI builders a shared open pipeline.
Key Takeaways
NVIDIA and Hugging Face announced on July 7, 2026 that Isaac GR00T 1.7 and the Isaac Teleop data-collection framework are being integrated into LeRobot, Hugging Face's open-source physical AI library, with a Cosmos 3 world-model integration to follow.
The partnership creates an end-to-end open pipeline: Isaac Teleop captures human demonstrations in interoperable formats, GR00T 1.7 is post-trained via LeRobot workflows, and Cosmos 3 will generate synthetic training data when robot time is scarce.
The collaboration connects NVIDIA's 3 million robotics developers with Hugging Face's 16 million AI builders, and adds Isaac Sim, Isaac Lab and Isaac Lab-Arena to LeRobot's Environment Hub.
The largest open physical AI dataset, with over 350,000 real and simulated trajectories and 57 million grasps, is already downloadable through LeRobot.
A Jetson Thor port for the open-source Reachy 2 humanoid enables edge deployment of vision-language-action (VLA) models.
Kaan Tınmaz
NVIDIA and Hugging Face on July 7, 2026 announced an expanded collaboration that brings the Isaac GR00T 1.7 open reasoning vision-language-action (VLA) model and the Isaac Teleop data-collection framework directly into LeRobot, Hugging Face's open-source library for physical AI. A Cosmos 3 frontier world-model integration is planned to follow, giving open-source developers a common pipeline from data collection to policy training and simulation.
An Open Pipeline for Robot Foundation Models
The move stitches together three previously fragmented layers of the humanoid stack. Isaac Teleop lets developers capture high-quality human demonstrations from external devices in interoperable formats and share the resulting datasets in LeRobot. Isaac GR00T 1.7 — billed as the first open, commercially viable robot foundation model — can then be post-trained through LeRobot workflows for new embodiments and tasks. Cosmos 3, coming soon, will let developers generate synthetic scenes and augment training data when physical robot time is scarce.
Connecting 19 Million Developers
NVIDIA said the partnership connects its 3 million robotics developers with Hugging Face's 16 million AI builders. The two companies are also plugging NVIDIA Isaac Sim, Isaac Lab and Isaac Lab-Arena into LeRobot's Environment Hub, and a Jetson Thor port for the open-source Reachy 2 humanoid enables edge VLA deployment. The largest open physical AI dataset — more than 350,000 real and simulated trajectories and 57 million grasps — is already downloadable through the same channel.
Why It Matters
Open-source robotics has lagged the LLM ecosystem in part because data, simulators and hardware were siloed. LeRobot's growing catalogue, paired with GR00T's benchmarked post-training path, gives startups a lower-cost route to production models — a critical bridge as humanoid manufacturers accelerate deployments alongside programs like NVIDIA's Cosmos 3 and moves such as Agility Robotics' SPAC public listing. The playbook also complements NVIDIA's Halos safety framework for physical AI, giving builders a shared cadence from data through deployment.
Reporting based on coverage from NVIDIA Blogs, RoboticsTomorrow and The AI Insider.
Isaac GR00T 1.7 is NVIDIA's open reasoning vision-language-action (VLA) model, billed as the first open, commercially viable robot foundation model. Through LeRobot workflows it can be post-trained for new robot embodiments and tasks.
What does the NVIDIA-Hugging Face collaboration include?
Announced July 7, 2026, it brings Isaac GR00T 1.7 and the Isaac Teleop data-collection framework into Hugging Face's LeRobot library, plugs Isaac Sim, Isaac Lab and Isaac Lab-Arena into LeRobot's Environment Hub, adds a Jetson Thor port for the Reachy 2 humanoid, and will later integrate the Cosmos 3 frontier world model.
What is Isaac Teleop used for?
Isaac Teleop lets developers capture high-quality human demonstrations from external devices in interoperable formats and share the resulting datasets in LeRobot for training robot policies.
Why does this partnership matter for open-source robotics?
Open-source robotics has lagged the LLM ecosystem because data, simulators and hardware were siloed. A shared pipeline from data collection to policy training and simulation, plus GR00T's benchmarked post-training path, gives startups a lower-cost route to production models.
