Nvidia has unveiled Cosmos 3, a frontier open model built for physical AI. The model is designed to help robots, autonomous vehicles, and vision AI systems simulate and predict real-world environments.
Physical AI refers to systems that operate in the physical world, as opposed to models that only process text or images. These systems need to understand and respond to their surroundings in real time. That requires the ability to simulate possible scenarios and predict what happens next.
What the model does
Cosmos 3 gives these systems a way to model the world. Instead of reacting blindly, they can anticipate changes. That could help a robot plan a path around obstacles, or help a self-driving car decide when to brake. The model's core capability is simulation and prediction, letting machines run through many possible futures before choosing an action.
Why openness matters
By releasing Cosmos 3 as an open model, Nvidia is making the technology available to a wider set of developers. Open models allow researchers and companies to inspect, adapt, and build on top of the underlying technology. That can speed up progress across the field, since teams don't have to start from scratch.
Robots and autonomous vehicles operate in messy, unpredictable environments. A single unexpected object can force a system to rethink its next move. Cosmos 3 aims to give these systems a head start by letting them simulate many possible outcomes before acting. That kind of foresight could help them handle surprises more gracefully.
The model's ability to simulate and predict real-world environments will be put to the test as developers begin using it in robots and autonomous vehicles.
