Cosmos 3 Super
Cosmos 3 Super is NVIDIA's language model. A large-scale world foundation model using a Mixture-of-Transformer architecture that couples a vision-language Reasoner with a video Generator for unified understanding and generation.
Capabilities
Input1/5
Text✓
Image·
Audio·
Video·
PDF·
Output1/5
Text✓
Image·
Audio·
Video·
Embedding·
Capabilities0/13
Reasoning·
Adaptive Reasoning·
Function Calling·
Parallel Function Calling·
Structured Outputs·
Native JSON Schema·
Web Search·
URL Context·
Computer Use·
Code Execution·
File Search·
Prompt Caching·
Assistant Prefill·
Versions
| Version | Released | Context | Input / 1M | Output / 1M | Status |
|---|---|---|---|---|---|
| Cosmos 3 Super | — | — | — | — | Current |
| Cosmos 3 Super Image2Video | — | — | — | — | Available |
| Cosmos 3 Super Text2Image Agentic | — | — | — | — | Available |
Other Models
| Model | Tier | Released | Context | Input / 1M | Output / 1M |
|---|---|---|---|---|---|
| Cosmos 3 Nano | Nano | — | — | — | — |
| Cosmos | — | — | — | — | — |
| Cosmos Reason1 7B | — | — | — | — | — |