NVIDIA is launching a suite of open-source AI technologies aimed at enhancing access and innovation across fields like language, robotics, and biology. This initiative includes the NVIDIA Nemotron family for AI reasoning, the Cosmos platform for physical AI, Isaac GR00T for robotics, and Clara for biomedical applications. By contributing over 650 open models and 250 datasets to Hugging Face, NVIDIA is making significant strides in democratizing AI resources for developers worldwide. Hugging Face's CEO, Clément Delangue, emphasized that these open models will promote transparency and accountability in AI development. The Nemotron family introduces several specialized models, including Nemotron Nano 3, which enhances reasoning for customer service and IT support, and Nemotron Safety Guard, which moderates content in multiple languages. NVIDIA also released new datasets focused on multimodal training and privacy-preserving synthetic data, along with tools like NeMo Data Designer for synthetic data generation. Companies like ServiceNow and Palantir are already leveraging these models to create agentic AI systems that transform traditional applications into more interactive and intelligent solutions. Updates to the Cosmos and Isaac GR00T models aim to improve robotic reasoning and world simulation. Cosmos Predict 2.5 can generate 30-second videos from a single frame, while Isaac GR00T N1.6 focuses on humanoid robot control. NVIDIA's open-source dataset for physical AI now includes 1,700 hours of multimodal driving data, attracting interest from firms like Amazon Robotics and Uber. In healthcare, the Clara family has introduced models like Clara CodonFM, which studies RNA rules to aid in therapy design, and Clara La-Proteina, which creates complex 3D protein structures. These developments position NVIDIA as a key player in advancing AI applications across various sectors.