A world model is an AI system that learns a predictive representation of how an environment changes over time. It can use observations, actions, or language to estimate future states and possible outcomes before an agent acts. World models support planning in model-based reinforcement learning, robotics, embodied AI, autonomous driving, and interactive simulation. This site catalogs these systems with model records, research papers, benchmark context, and source links. Start with the world models database, compare systems on the leaderboard, or read the Performance Index methodology.
Short extractable summary preserved directly in static HTML.
Editorial provenance and refresh policy preserved directly in static HTML.
Published by world-models.io editorial board.
Lead editor world-models.io editorial board - Event review and source normalization.
This hub monitors verified primary sources, normalizes event data, and publishes reviewed summaries linked to the broader knowledge graph.
Each editorial page is assembled from primary sources, normalized into extractable summaries, checked for factual drift, and reviewed before publication or major refreshes. Last reviewed: 2026-09-27.
Pages are refreshed when a new paper, benchmark, release, architecture update, or stronger primary source materially changes the answer a reader or AI system should retrieve.
Each page links back to relevant primary sources and keeps a stable canonical URL so readers can verify claims, trace context, and reference the most up-to-date version. See the editorial policy.
Static overview of the release archive.
The timeline tracks published world model papers, product releases, benchmark updates, and lab announcements in chronological order.
This static snapshot exposes 21 timeline events before JavaScript runs, so crawlers can discover the event archive and detail pages directly from HTML.
Most recent events rendered directly in static HTML.
| Event | Source | Type | Published | Summary |
|---|---|---|---|---|
| World-models.io launches a read-only MCP connector for AI clients | world-models.io Editorial | editorial_publish | 2026-10-30T15:00:00Z | World-models. io launches a read-only MCP connector for AI clients. |
| Runway introduces Praxis-1, an open-weight world action model for robotics | Runway Research | model_release | 2026-09-30T00:00:00Z | Runway introduces Praxis-1, a world action model that transfers large-scale video pretraining into robot control. |
| GAVEL proposes graph world models for verified long-horizon LLM planning | arXiv | paper_published | 2026-09-16T18:25:58Z | GAVEL presents an explicit graph world model for checking and repairing long-horizon plans proposed by language models. |
| PSG-JEPA asks whether forward prediction is enough for physically grounded world models | arXiv | paper_published | 2026-08-07T04:44:16Z | PSG-JEPA studies whether forward prediction alone gives a JEPA world model the physically meaningful state needed for control. |
| A practical taxonomy for AI world models, published by the editorial team | world-models.io Editorial | editorial_publish | 2026-07-14T10:00:00Z | A new editorial taxonomy compares AI world models across domain, function, internal representation, temporal horizon, action conditioning, evaluation type, availability. |
| Physically Viable World Models make the case for query-conditioned embodied AI | arXiv | paper_published | 2026-05-28T20:18:22Z | This position paper argues that embodied world models should answer intervention queries about the consequences of actions, not merely generate visually plausible future. |
| Google DeepMind unveils Genie 3: real-time interactive world model at 720p/24fps | Google DeepMind | model_release | 2026-04-22T14:00:00Z | DeepMind announces Genie 3, a real-time foundation world model generating interactive 3D environments at 720p and 24 fps from text prompts. |
| NVIDIA Cosmos-Predict2 released: improved physical consistency for robotics simulation | NVIDIA Research | model_release | 2026-04-18T16:00:00Z | NVIDIA Research releases Cosmos-Predict2, a second-generation diffusion world model with stronger physical consistency for robot policy pretraining. |
| Wayve publishes GAIA-2 technical report: scalable driving world model | arXiv | paper_published | 2026-04-12T09:00:00Z | Wayve releases the GAIA-2 technical report detailing a multi-camera, multi-condition driving world model trained on 4,700 hours of urban driving. |
| Meta FAIR releases V-JEPA 2 checkpoints under non-commercial license | Meta FAIR | model_release | 2026-04-08T15:00:00Z | Meta FAIR publishes V-JEPA 2 model checkpoints (ViT-L, ViT-H, ViT-g) on Hugging Face under a research-only license. |
| world-models.io launches 'Frontier World Models 2026' editorial collection | world-models.io Editorial | collection_update | 2026-04-05T10:00:00Z | A new curated collection on world-models. |
| 1X Technologies and DeepMind announce joint NEO world model for humanoid manipulation | Google DeepMind | model_release | 2026-03-28T13:00:00Z | 1X Technologies and Google DeepMind announce a joint research effort on NEO, a world model for humanoid manipulation trained on teleoperation data. Initial results show 2. |
Canonical crawlable directory of timeline events. JavaScript filters can enhance this same archive for interactive readers.
Primary references and official sources surfaced directly in static HTML for crawlers and no-JS readers.