AI BriefingNVIDIA Unveils Cosmos 3 World Model
Key point
NVIDIA has unveiled Cosmos 3, an omnimodal world model for physical AI research.
Details
NVIDIA has released an omnimodal world model series named Cosmos 3 on Hugging Face. This model can combine text, image, video, and action trajectory inputs to generate high-quality dynamic video, images, audio, and action commands.
The main model configurations are as follows:
- Nano: 16B parameters
- Super: 64B parameters
Cosmos 3 is designed as a core building block for a wide range of Physical AI research and applications, including world understanding, world generation, simulation, and embodied policy learning.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Privatemode Benchmarks GLM-5.3-Flash as Vision-Capable Decision Model with Parity to Jev but Higher Cost