NVIDIA Unveils Cosmos Models and Infrastructure for Robotics

NVIDIA has announced new Cosmos world models and infrastructure for robotics at the SIGGRAPH conference, announced in a press release. The highlight is Cosmos Reason, a 7-billion-parameter vision language model designed for physical AI applications and robotics.

Cosmos Reason enables robots and AI agents to perform reasoning tasks by understanding memory and physics, which can be applied in data curation, robot planning, and video analytics. Additionally, NVIDIA introduced Cosmos Transfer-2, which accelerates synthetic data generation from 3D simulation scenes, and a distilled version optimized for speed.

The company also launched new neural reconstruction libraries for simulating real-world environments in 3D using sensor data. These capabilities are integrated into the open-source simulator CARLA and the Omniverse software development kit. New servers, such as the NVIDIA RTX Pro Blackwell Server and NVIDIA DGX Cloud, support these advancements by providing a unified architecture for robotic development workloads.