A source-checked directory of 24 AI systems that create, simulate, reconstruct, or reason about worlds.
This repository is a compact reading layer, not a leaderboard. It separates interactive world models, persistent 3D creation systems, physical-AI models, and adjacent research so that similarly marketed projects are not treated as interchangeable.
Last verified: September 4, 2026
- Browse all 24 systems
- Read the classification and source policy
- Open the Pages edition
- Explore the full World Models Watch directory
The term covers systems that learn or generate a representation of how an environment looks, changes, or responds to actions. Some render explorable video or 3D scenes. Others predict physical outcomes for robots. A few entries are included as boundary cases because they clarify what a world model is not.
Each entry records the maker, category, release date, current public status, a short original summary, a boundary note, and at least one primary source.
This repository contains editorial summaries only. It does not include model weights, product code, third-party logos, benchmark reproductions, private research notes, or media copied from the projects it describes.
For deeper model profiles, comparisons, release tracking, and practical world generation guides, visit World Models Watch.
Original text in this repository is available under CC BY 4.0. Names, trademarks, linked media, and third-party materials remain the property of their respective owners. See NOTICE.md.