On Monday, NVIDIA unveiled a set of New World AI models, libraries and other infrastructure for robotics developers. Most notably, COSMOS Reason, a 7 billion parameters “inference” vision language model for physical AI applications and robots.
Additionally, participating in existing batches of Cosmos World models are Cosmos Transfer-2, which can accelerate the generation of synthetic data from 3D simulation scenes or spatial control inputs, and distilled versions of COSMOS transfers that are more optimized at speed.
During its presentation at the Siggraph Conference Monday, Nvidia noted that these models are intended to be used to create synthetic text, image and video datasets for training robots and AI agents.
According to Nvidia, for Cosmos reasons, robots and AI agents can “infer” thanks to their memory and understanding of physics. The company says it can be used for data curation, robot planning and video analysis.
The company has also announced a new neural reconstruction library. This includes rendering techniques that allow developers to simulate the real world of 3D using sensor data. This rendering feature is also integrated into the popular developer platform, Carla open source simulator. There is also an update to the Omniverse Software Development Kit.
There is also a new server for robotics workflows. While NVIDIA RTX Pro Blackwell Server offers a single architecture for robot development workloads, NVIDIA DGX Cloud is a cloud-based management platform.
These announcements come as the semiconductor giants are pushing further into robotics as they look beyond AI data centers to the next major use case for AI GPUs.
TechCrunch Events
San Francisco
|
October 27th-29th, 2025
We are constantly trying to evolve and you can help us by providing insights about your perspective and feedback to TechCrunch! Fill in this survey and let us know how we are doing.
Source link
