Change the world with words
You can steer a running simulation in plain language. Google’s examples include changing the weather, moving the camera, and adding a new person or object to the scene.
Genie 4: what it is, demos & API status
Last updated:
Genie 4 is a world model from Google DeepMind, announced on October 27, 2026. Instead of making a video you watch, it generates a world you can move through and change while it runs, with sound and characters that talk back. Right now only trusted testers can use it. Google says it is coming next to Project Genie for Ultra subscribers and to early partners of the Gemini API.
At a glance
- Made by
- Google DeepMind
- Announced
- Oct 27, 2026
- Status
- Trusted-tester preview
- Public app
- Coming to Project Genie
- API
- Early partners only
What is Genie 4?
Genie 4 is the fourth generation of Google DeepMind’s Genie world models. A world model learns how an environment behaves, so it can predict what happens next when you act in it. In practice, that means you describe a scene, step into it, and the model keeps generating what you see and hear as you move and give instructions. DeepMind says Genie 4 builds on capabilities from the Gemini family of models: Gemini helps with how the world looks and behaves, and can watch the scene and break a goal into steps. Google’s stated reasons for building it range from training environments for robots and AI agents to new kinds of interactive experiences. Testers in creative fields have used it for pre-visualization and early ideation. Genie 4 follows Genie 3, which DeepMind announced in August 2025 and which powers today’s Project Genie app.
What Genie 4 can do
More realistic physics and people
DeepMind says Genie 4 improves how physical interactions and human movement look, including how characters react to what is around them. These are Google’s own claims, not independent test results.
Built-in sound and speech
Genie 4 is natively multimodal. It generates ambient sound for the world, and characters can speak and react to your inputs or to events in the scene in real time.
Multi-step tasks with Gemini
Gemini can observe the simulation, break a goal into a sequence of steps and guide what happens next. DeepMind’s demo shows this with a humanoid robot playing checkers.
Get notified when the Genie 4 API opens.
One email when the API opens. Nothing before that.
No spam · Unsubscribe anytime Privacy
Official Genie 4 demo
This is Google DeepMind’s official announcement video. DeepMind notes that the results shown are illustrative and the sequences are shortened, so expect real sessions to look rougher in places.
Can I use Genie 4 now?
Not unless you are one of Google’s trusted testers. There is no public app or API for Genie 4 yet. You have three realistic options today. First, join Google’s official waitlist for Genie 4 in Project Genie, which DeepMind links from its announcement. Second, if you have a Google AI Ultra subscription and live in a supported country, you can use Project Genie now; it currently runs on Genie 3. Third, try one of the other world models that are open today, such as World Labs’ Marble or the open-source HunyuanWorld and Matrix-Game.
Genie release timeline
- Aug 5, 2025
Genie 3 announced
Limited research preview: 720p at 24 frames per second, consistent for a few minutes.
- Jan 29, 2026
Project Genie launched
Web app built on Genie 3, for Google AI Ultra subscribers in the US aged 18+.
- Oct 27, 2026
Genie 4 announced
Research preview for trusted testers. Project Genie and Gemini API partners next.
- Not announced
Public Genie 4 access
No date yet for the Project Genie rollout or a public API.
Genie 4 vs Genie 3 at a glance
Project Genie launched on Genie 3, together with Nano Banana Pro and Gemini. Google says Genie 4 is coming to Project Genie but has not said when every subscriber will get it.
| Genie 3 | Genie 4 | |
|---|---|---|
| Made by | Google DeepMind | Google DeepMind |
| Announced | Aug 5, 2025 | Oct 27, 2026 |
| Where to use it | Project Genie (Ultra) | Coming to Project Genie |
| Access at launch | Limited research preview | Trusted-tester preview |
| Public API | Not announced | Early partners announced |
Genie 4 limitations
DeepMind is upfront that Genie 4 is an early, experimental model. Its model page lists several limits. Visual details and object permanence can shift or drift over time, so something you walked past may look different when you come back. It cannot simulate real-world places with perfect accuracy. Clear, readable text usually appears only when it is part of your world description. Simulating other agents and complex interactions is still an open research problem, and real-time audio accuracy is still being improved. Sessions are designed for a few minutes of continuous interaction, not long play sessions. None of this comes from our own testing; we have not used Genie 4. It is a summary of what Google itself publishes.
FAQ
What is Genie 4?
Genie 4 is a world model from Google DeepMind, announced on October 27, 2026. It generates interactive worlds you can move through and change in real time with language, including sound and speech.
Is Genie 4 free?
Google has not announced a price or a free tier for Genie 4. Today’s Project Genie app, which runs on Genie 3, requires a Google AI Ultra subscription.
When will Genie 4 be released?
The research preview for trusted testers began on October 27, 2026. Google says Genie 4 is coming to Project Genie and early Gemini API partners next, but has not given a date.
Is there a Genie 4 API?
Not publicly. Google has named early access partners of the Gemini API and Google AI Studio as an upcoming route, with no public price or release date yet.
Does Project Genie use Genie 4?
Project Genie launched on Genie 3. Google says Genie 4 is coming to Project Genie but has not confirmed that the rollout is complete.
Get notified when the Genie 4 API opens.
One email when the API opens. Nothing before that.
No spam · Unsubscribe anytime Privacy