Nvidia has released a new artificial intelligence model built for robots and other machines that operate in the physical world.
According to The Information, the company on Wednesday unveiled Cosmos 3 Edge, described as a small, open-source model for "physical" AI — systems designed to work in the real world rather than only on screens. The Information reports the model has just 4 billion parameters, a measure of an AI system's size, making it compact enough to run on lower-powered hardware. (The Information's report was cut off before it detailed exactly where the model can run.)
Cosmos is part of a broader push by Nvidia into robotics and machine perception. GamesBeat reports that Nvidia's Metropolis platform is being used to speed up the creation of Cosmos-powered "vision AI agents" — software that can interpret what cameras and sensors see.
The effort also has an international dimension. According to Chosunbiz, Nvidia is deepening its ties with Japan and targeting manufacturing and robotics applications using Cosmos.
The word "open-source" is notable here: it signals that developers and companies outside Nvidia can access and build on the model directly, rather than paying for a closed, proprietary system. Combined with its small size, that could make it easier for robot makers and factories to adopt.
The sources do not specify pricing, licensing terms, benchmark performance, or a full list of supported devices, so key details about how Cosmos 3 Edge compares to rivals remain unclear from the available reporting.
Why it matters: "Physical AI" is widely seen as the next frontier for the technology, and a small, open model from a company as central to the AI industry as Nvidia could lower the barrier for building smarter robots and factory systems.