NVIDIA Launches Cosmos 3, World's First Fully Open Omnimodel, Alongside Global Cosmos Coalition
Summary
NVIDIA launches Cosmos 3, the world's first fully open omnimodel capable of natively processing text, images, video, sound, and actions for robotics and autonomous vehicles, while simultaneously unveiling the global Cosmos Coalition to accelerate open world model development across industries.
Key Points
- NVIDIA launches Cosmos 3, the world's first fully open omnimodel built on a mixture-of-transformers architecture, capable of natively understanding and generating text, images, video, ambient sound, and actions for physical AI applications including robotics and autonomous vehicles.
- Cosmos 3 tops multiple physical AI benchmarks and is available in three variants — Cosmos 3 Super, Cosmos 3 Nano, and the upcoming Cosmos 3 Edge — reducing physical AI training and evaluation cycles from months to days.
- NVIDIA simultaneously launches the Cosmos Coalition, a global collaboration with founding members including Agile Robots, Black Forest Labs, Runway, and Skild AI, aimed at accelerating open world model development across industries using shared infrastructure and research.