NVIDIA released four new open-weight Cosmos 3 Nano checkpoints. Cosmos3-Nano-Transfer-Auto generates multi-view driving data with varied weather and lighting. Cosmos3-Nano-Sim-Auto streams synchronized camera and LiDAR observations for driving simulation. Cosmos3-Nano-Sim-Transfer turns simulator controls and text prompts into photorealistic video. Cosmos3-Nano-Sim-Bimanual predicts video from robot or camera actions. All models have 16B trainable parameters. They use the OpenMDW1.1 license for commercial and non-commercial use. Developers can adapt the models to their own sensor rigs and domains. Streaming generation supports fast, interactive rollouts. The models are available on Hugging Face.
What changed
Cosmos 3 Nano had no open-weight checkpoints for interactive world simulation or multi-sensor driving scenes.
What it unlocks
Generating interactive world simulations and multi-sensor driving scenes with open weights.
- 4 new open-weight checkpoints
- 16B trainable parameters per model
- up to 11 camera views
- 4 RGB frames per latent time step
What you need to act on it
- Linux
- an NVIDIA GPU
- Docker
- NVIDIA Container Toolkit
Sources