About this model or application
Cosmos3 models process text, visual input, audio and action trajectories, and generate multiple modalities for world simulation. This is separate from a conventional chat model.
Family & other versions2 sources
Background for the family or other versions. These sources are counted separately from evidence for this release.
Model cards & docsOfficial
Cosmos3-Super official documentationFamily & other versions
Official model specifications and deployment information.
Model cards & docsOfficial
Cosmos3-Nano official documentationFamily & other versions
Official model specifications and deployment information.
Model family catalog2
Current models
Official catalogCosmos3-SuperAvailable256K ContextOpen weights
- Model ID
nvidia/Cosmos3-Super- Released
- 2026-05-31
- Availability
- Available
- Input
- Text · Image · Video · Audio · Action
- Output
- Text · Image · Video · Audio · Action
- Context
- 256,000 tokens
- Parameters
- 64B
- Deployment / access
- Open weights · Local
- Weights
- Open weights
- License
- openmdw1.1-license
Physical-AI world simulation. Audio is embedded in video; action is a trajectory/list, not a 3D asset.
Cosmos3-NanoAvailable256K ContextOpen weights
- Model ID
nvidia/Cosmos3-Nano- Released
- 2026-05-31
- Availability
- Available
- Input
- Text · Image · Video · Audio · Action
- Output
- Text · Image · Video · Audio · Action
- Context
- 256,000 tokens
- Parameters
- 16B
- Deployment / access
- Open weights · Local
- Weights
- Open weights
- License
- openmdw1.1-license
Physical-AI world simulation. Audio is embedded in video; action is a trajectory/list, not a 3D asset.