# Neural World Models

Navigation World Models predict what a robot's camera will see after a navigation action, and can plan a route by imagining trajectories before taking them. Code and weights are for non-commercial use.

Organization: Meta (FAIR). Released: December 2024. Kind: Explorable world.

## What you control

- Past camera frames
- Navigation actions

## License

Weights: Available (https://huggingface.co/facebook/nwm). Declared license: CC BY-NC 4.0 (per README).
No. Hugging Face metadata says CC BY 4.0, but the project README licenses code and weights under CC BY-NC 4.0. The stricter, non-commercial reading is shown.

## Real-time

No real-time claim.

## How to run it

1. Clone facebookresearch/nwm and install the packages listed in the README.
2. Download the pretrained model.
3. Use isolated_nwm_infer.py to predict the next view after an action or to follow a recorded trajectory, and isolated_nwm_eval.py to score the results.

## Where to try

No hosted destination verified.

## Sources

- [Project source](https://github.com/facebookresearch/nwm)
- [Weights and model card](https://huggingface.co/facebook/nwm)

## Questions

### Is Neural World Models open source?

Neural World Models publishes downloadable weights under CC BY-NC 4.0 (per README). Read the license terms before you build on it.

### Can I use Neural World Models commercially?

No. Hugging Face metadata says CC BY 4.0, but the project README licenses code and weights under CC BY-NC 4.0. The stricter, non-commercial reading is shown.

### Does Neural World Models run in real time?

The cited sources do not claim real-time operation.

### Where can I try Neural World Models?

No hosted version was verified; run the published weights yourself.

### What does Neural World Models take as input?

You give it past camera frames and navigation actions.

### How do I run Neural World Models?

Following the official instructions: Clone facebookresearch/nwm and install the packages listed in the README. Download the pretrained model. Use isolated_nwm_infer.py to predict the next view after an action or to follow a recorded trajectory, and isolated_nwm_eval.py to score the results.

### What are alternatives to Neural World Models?

Similar models in this directory include ABot-World (open weights, permissive license), AlayaWorld (open weights, custom terms) and DreamX-World (open weights, permissive license).

Checked 2026-10-23. Corrections: https://github.com/windflow-labs/world-model-directory