VarwoGoogle Genie 4: World Model, Access and Preview Limits
Understand what Genie 4 simulates, how it uses Gemini and what the official research preview currently makes available.
Genie 4 is Google DeepMind's world model for interactive simulation. It generates a scene that responds as a user navigates and steers it, rather than only producing a fixed video clip. DeepMind's October 27 announcement describes native environmental sound, music and dialogue alongside the generated world.
Can you use it now?
As checked October 31, 2026, the announcement describes a research preview for trusted testers. It says access is coming to Google AI Ultra subscribers through Project Genie and early-access partners of the Gemini API and Google AI Studio. Use the waitlist linked by DeepMind. This does not establish general API availability, an unrestricted download, a confirmed admission date or a public price.
Why searches mention Gemini
DeepMind says Genie 4 incorporates capabilities from the Gemini family for visual grounding and scene coherence. That relationship does not make every Gemini chat or video-generation endpoint a Genie 4 session. Gemini 4 Carbon is a separate model topic.
| What you need | What to look for |
|---|---|
| Explore and change an interactive scene | Genie preview eligibility and controls |
| Export a predetermined video | A video-generation workflow and its export terms |
| Train an agent in simulation | Research access, repeatability and evaluation conditions |
Read demonstrations correctly
The announcement's recordings show interactions selected by the provider. They are reference demonstrations, not a playable Varwo demo or an independent reliability test. Before adopting a future preview, check session duration, persistence, failure handling and export permissions against the actual account offered access.
Varwo reviewed the official announcement; we did not access a Genie 4 preview, generate a world or test its physical accuracy.