Nvidia unveils new Cosmos world models, infra for robotics and physical uses

On Monday, Nvidia unveiled a set of New World AI Model, Library and other infrastructure for Robotics developers, which is a significant reason for physical AI app and 7-billion-parameter “logic” Vision Language model for robots.

Also joined the existing batch Cosmos World Model The cosmos transfer -2, which can accelerate synthetic data generation from 3D simulation scenes or spatial control inputs, and a subtile version of cosmos transfer that is more favorable for speed.

When announced at the Siggraph conference on Monday, the NVIDIA mentions that these models are used to create synthetic text, images and video data sets for training of robots and AI agents.

Nvidia per cosmos because robots and AI agents will thank the “logic” for the memory of the agents and physics, which lets it be “a model of a plan to argue what any figure agent can take the next action.” The agency says it can be used for data creation, robot plan and video analysis.

The company also unveiled the new neural reconstruction library, one of which includes one for rendering technology that allows developers to imitate the real world in 3D using sensor data. This rendering capacity is also integrated with a popular developer platform Open Source Simulator Carla. Even the Omonvers Software Development Kit has an update.

There is also a new server for robotics workflies. Nvidia RTX Pro Blackwell Servers provides a single architecture for robotic development work stress, while on the other hand NVDia DGAX Cloud is a cloud-based management platform.

These announcements have come up since the semiconductor giant is pressing even more on robotics because it is out of AI data centers and shows the next major use for its AI GPUs.