Nvidia has introduced a new set of world-class AI models, libraries, and infrastructure designed specifically for robotics developers. The most notable model is Cosmos Reason, a 7-billion-parameter “reasoning” vision language model that enables robots to understand physical interactions and make decisions.
The Cosmos series includes additional models, such as Cosmos Transfer-2, which accelerates synthetic data generation from 3D simulation scenes, and a distilled version of Cosmos Transfers for faster processing. These models are intended for training robots and AI agents with realistic datasets, including text, images, and video.
Cosmos Reason allows robots to “reason” by leveraging its memory and physics understanding, making it suitable for tasks such as data curation, robot planning, and video analytics. The company also unveiled new neural reconstruction libraries, which include a rendering technique that simulates the real world in 3D using sensor data.
Additionally, Nvidia has introduced new servers for robotics workflows, including the Nvidia RTX Pro Blackwell Server and Nvidia DGX Cloud, a cloud-based management platform. These announcements mark Nvidia’s continued push into robotics and its exploration of AI applications beyond traditional AI data centers.
The news comes as Nvidia looks to expand its presence in the robotics market and develop new use cases for its AI GPUs. The company is also set to participate in TechCrunch Disrupt 2025, alongside other prominent tech and VC heavyweights such as Netflix, ElevenLabs, Wayve, Sequoia Capital, and Elad Gil.
Source: https://techcrunch.com/2025/08/11/nvidia-unveils-new-cosmos-world-models-other-infra-for-physical-applications-of-ai