archive
· today in ai · 2026-08-06
NVIDIA pushes world models for robotics
Archive item — written before sources were shown.
NVIDIA detailed Cosmos 3, combining vision reasoning, world generation, and action prediction to train robotics and autonomous-vehicle systems in simulation.
NVIDIA’s Cosmos 3 combines vision-based reasoning, generative world modeling, and action prediction across multiple model sizes, letting teams simulate how physical environments behave to train and specialize robotics and autonomous-vehicle systems without relying solely on real-world data collection.
sources
- 01Into the Omniverse: How Open World Models Push the Frontier of Physical AIblogs.nvidia.com · primary
Rami Steitieh