Tasks and inputs
Cosmos3-Nano-Sim-Auto is listed for Text-to-video, Video-to-video. It accepts text, video, image and produces video. Use the publisher’s input format and preprocessing for your chosen deployment.
SyncDev model directory · Profile updated Nov 19, 2026
Cosmos3-Nano-Sim-Auto is NVIDIA's open causal driving simulator: real-time multi-camera video and LiDAR that react to an AV policy's planned path.
The specification score summarizes recorded features and limits. It is not a benchmark result.
Pricing conditions: Open weights; no hosted API price. GPU time is the cost, so confirm throughput on your own hardware.
Recorded modality support shows what this model can accept and produce. An undocumented modality is shown as unknown rather than assumed to be unsupported.
Accepts
Produces
A driving policy that only ever sees recorded video never learns what happens after its own mistakes. Cosmos3-Nano-Sim-Auto closes that loop. The policy plans a trajectory, a World Scenario Map renderer turns it into scene controls, and the model generates what every camera and the LiDAR would observe next, chunk by chunk.
Release date: 20 October 2026. Its parent is the bidirectional Cosmos3-Nano-Transfer-Auto, and the whole point of the distillation is that this version can generate the next stretch of every sensor stream before the planner has to decide again.
First, Cosmos3-Nano was post-trained on multi-sensor driving data including LiDAR, producing Transfer-Auto. Second, that model was causally distilled into this few-step generator. The architecture stays a 16-billion-parameter Mixture-of-Transformers.
Outputs are MP4 or streamed frames per view and LiDAR range-intensity maps convertible to point clouds.
| Constraint | Value |
|---|---|
| Views | Up to 11, Hyperion 8.1 rig |
| LiDAR conditioning | One initial sweep only; longer measured prefixes unsupported |
| Resolution | 480p or 720p |
| Playback | 10–30 fps across selected views |
| Tested GPUs | H100, H20, H200, B200, B300, GB200, GB300 among others |
| Licence | OpenMDW-1.1 |
For a custom rig, the documented path is to post-train Transfer-Auto on your sensors first and then repeat the causal distillation.
Run three checks before using it to score a planner:
Any failure matters. Closed-loop scores built on a drifting world will mislead you.
Related NVIDIA models include the edge driving model Alpamayo 2 Edge; more video models are in the video-to-video directory.
Compare the model on the work you need to complete.
Cosmos3-Nano-Sim-Auto is listed for Text-to-video, Video-to-video. It accepts text, video, image and produces video. Use the publisher’s input format and preprocessing for your chosen deployment.
Choose duration, resolution and frame rate together. Compare the cost of an accepted clip, including repeat generations.
The weights are downloadable under OpenMDW-1.1. Check runtime support and size the complete model and working context for your expected concurrency.
These records share a directory category. Check the source card, output type and deployment requirements before treating a model as a replacement.
NVIDIA
Cosmos3-Nano-Sim-Transfer is NVIDIA's open 16B model that renders simulator depth, segmentation or edge maps into photorealistic robot video in real time.
NVIDIA
Cosmos3-Nano-Transfer-Auto is NVIDIA's open 16B model that generates synchronised multi-camera driving video and LiDAR from map controls and a prompt.
skywork
Matrix-Game 4.0 is Skywork's open set of 5B real-time video renderers that turn a running game's frames into clean styled video, per scene.
Tencent
WeVisCut is Tencent's 9B tool-calling agent that edits narration scripts into multi-minute videos through 12 editing tools and an editable Studio timeline.
skywork
SkyReels-V3-V2V-14B is Skywork's 14B video-extension model that continues a clip for 5-30 seconds or cuts to a new shot, at 720p from a text instruction.
AInVFX
AInVFX/SeedVR2_comfyUI is a source-linked Hugging Face repository assigned to Video-to-video. Its recorded task interface accepts video and produces video.