NVIDIA Cosmos is NVIDIA's open set of world models, AI that simulates how scenes look and change, used to help robots and self-driving cars learn.
In plain words
NVIDIA's goal is richer simulated worlds where robots and driving systems can be trained and tested, and it says Cosmos 3 can speed up how robots learn. The Cosmos 3 family can reason about a scene, simulate it and generate actions. In November 2026 NVIDIA added four open-weight Cosmos 3 Nano models: two create driving scenes, including matching camera and LiDAR views, one turns simulator controls into realistic video, and one predicts video from a robot's actions.
A quick example
Imagine a self-driving team that has footage of a road only on sunny days. They use a world model to produce the same drive in rain and at night, then test how their driving software reacts.
Where you'll meet it
- Making training data for robots and self-driving cars.
- Testing how a driving or robot policy responds in simulated scenes.
Related
Sources
- DocsNVIDIA Cosmos: Physical AI Development Platform, NVIDIA · read 5 Nov 2026
- DocsNew NVIDIA Cosmos 3 Open Models for Interactive World Simulation, NVIDIA on Hugging Face · read 5 Nov 2026
Checked by WhatIsAI against the sources above · 5 Nov 2026. Spotted an error? Report it.