NVIDIA has announced a major expansion of its open-source AI initiative, introducing new open models, datasets, and tools across language, biology, and robotics to accelerate innovation. These contributions, part of the NVIDIA Nemotron family for AI reasoning, the NVIDIA Cosmos platform for physical AI, NVIDIA Isaac GR00T for robotics, and NVIDIA Clara for biomedical AI, are designed to empower developers worldwide and strengthen U.S. technological leadership. The new open resources are being made available through Hugging Face, where NVIDIA has become one of the platform’s top contributors with over 650 open models and 250 open datasets. This commitment supports a transparent, collaborative, and responsible AI ecosystem. NVIDIA Nemotron brings ultra-efficient reasoning to specialized AI agents, enabling them to perceive, retrieve, generate, and reason using multiple models. The latest release includes new open-source datasets focused on multimodal training, multilingual personas, and privacy-preserving synthetic personal information. Developers also gain access to enhanced NeMo tools such as NeMo Data Designer for synthetic data generation and NeMo-RL for advanced post-training and reinforcement learning, giving greater control over model customization. Leading software companies are already leveraging Nemotron to build agentic AI systems. ServiceNow used Nemotron models to develop Apriel 2, a multimodal reasoning model that ranks at the top of leaderboards for document intelligence and customer service. Bill McDermott, chairman and CEO of ServiceNow, praised NVIDIA’s role in enabling secure, domain-specific AI that scales and transforms enterprise workflows. In robotics and physical AI, NVIDIA unveiled updates to its Cosmos world foundation models and Isaac GR00T robot foundation models. These enable robots to learn humanlike reasoning and adapt to complex environments. The company also released the world’s largest open-source dataset for physical AI, featuring 1,700 hours of multimodal driving data from the U.S. and Europe, along with GR00T training data that has become one of the most-downloaded datasets on Hugging Face. Companies including Agility Robotics, Amazon Robotics, Figure AI, Skild AI, Milestone Systems, and Uber are using these models to generate synthetic data, train new robot behaviors, and deploy physical AI agents at scale. In healthcare and life sciences, NVIDIA Clara now includes new open models for scientific discovery, medical image analysis, and drug development. These tools accelerate research and improve diagnostics by leveraging AI trained on real-world biomedical data. Developers can access NVIDIA’s open models through build.nvidia.com, Hugging Face, OpenRouter, and Microsoft Azure AI Foundry, with upcoming availability on Google Vertex AI. NVIDIA Clara and Isaac GR00T are available on Hugging Face, and all models are offered as NVIDIA NIM microservices for secure, scalable deployment on DGX Cloud or any NVIDIA-accelerated infrastructure. The move underscores NVIDIA’s vision of open collaboration to drive the next wave of AI innovation across industries. More details can be found in the keynote from NVIDIA GTC Washington, D.C., delivered by CEO Jensen Huang.