NVIDIA and Hugging Face have formally integrated key artificial intelligence tools into LeRobot, Hugging Face’s open-source robotics library, to accelerate the development of humanoid and general-purpose robots. This partnership introduces the NVIDIA Isaac GR00T 1.7 reasoning vision-language-action model and the Isaac Teleop data collection framework directly into the LeRobot ecosystem, with the NVIDIA Cosmos 3 physical AI foundation model scheduled for addition in the near term. The collaboration aims to dismantle the fragmented, resource-intensive barriers that have historically constrained physical AI innovation by providing developers with a unified, open-source pipeline for end-to-end robotics development. LeRobot serves as a standardized platform for training, running, and distributing robot datasets, models, and operational workflows. By embedding NVIDIA’s capabilities into this open-source environment, the partnership connects NVIDIA’s network of three million robotics developers with Hugging Face’s community of 16 million artificial intelligence engineers. The integration begins with Isaac Teleop, an open-source framework that enables developers to capture high-quality human robot demonstrations through external devices. These demonstrations are recorded in standardized, interoperable formats, allowing teams to expand, validate, and share datasets directly within LeRobot. Following data collection, developers can utilize Isaac GR00T 1.7, recognized as the first commercially viable and openly accessible robot foundation model. This architecture simplifies post-training and deployment workflows, allowing engineers to efficiently fine-tune foundation models for diverse robot embodiments and specific task benchmarks. Looking ahead, the upcoming integration of Cosmos 3 will address critical data limitations by serving as a frontier world model for physical AI. This system will enable developers to synthesize and augment training datasets, simulate complex operational environments, and refine policy algorithms when real-world data collection is cost-prohibitive or impractical. Thomas Wolf, co-founder and chief science officer at Hugging Face, emphasized that open-source ecosystems are essential for translating advanced research into adaptable, community-driven technology. The unified workflow ensures that robotics professionals can train, evaluate, and deploy models transparently while contributing to a shared knowledge base. These integrations complement existing NVIDIA robotics resources already linked to LeRobot to support the complete development lifecycle. By standardizing data collection, model training, and policy deployment within a single open-source framework, NVIDIA and Hugging Face are establishing a scalable infrastructure for the next generation of physical AI. Developers seeking to implement these tools can access detailed integration workflows and documentation through the LeRobot platform.