NVIDIA reveals new Cosmos World models, and Infra for robots and physical uses

On Monday, NVIDIA has unveiled a set of new models and libraries of the world and other infrastructure of robotics developers, the most prominent of which is the reasons for being Cosmos, a 7 -billion “logical” vision language model for physical AI applications and robots.

Also join the current batch Cosmos World Models It is Cosmos Transfer-2, which can accelerate the generation of artificial data from 3D simulation scenes or spatial control inputs, and a cosmos more cosmos transfers.

During its announcement at the Siggraph conference on Monday, NVIDIA noticed that these models are supposed to be used to create artificial image and video data collections to train robots and artificial intelligence agents.

Cosmos Reaup, for all NVIDIA, of robots and artificial intelligence agents “mind” thanks to its memory and understanding of physics, allows it to “be a model for planning to cause steps that the embodied worker may take after that.” The company says it can be used to activate data, robot planning and video analyzes.

The company also unveiled new nervous reconstruction libraries, which includes one to present a introduction technology that allows developers to simulate the real world in 3D using sensor data. This application capacity is also combined into an open source simulation Carla, a famous developer platform. There is even an update of the Omniver’s software development group.

There are new servers for robots function as well. NVIDIA RTX Pro Blackwell server provides a single structure for the burdens of automated development work, while NVIDIA DGX Cloud is a cloud -based management platform.

These ads come at a time when the semiconductor giant pushes robots because it looks forward to the state of the next large use of GPUS AI after artificial intelligence centers.