Google Genie 4 is the world model that turns a text prompt into a playable, explorable world. Google first gave it to AI Ultra subscribers on 18 November 2026, then opened it to developers through the Gemini API a day later. It is the first time developers can build with a real-time world model from Google.
This guide explains what Google Genie 4 is, how much it costs, who can use it, and what its limits are.
Google Genie 4: Quick Facts
- What it is: A real-time interactive world model from Google DeepMind
- Announced for subscribers: 18 November 2026, in Project Genie for Google AI Ultra
- Opened to developers: 19 November 2026, through the Gemini API and Google AI Studio
- Model code: models/genie-4, through the Interactions API
- Video quality: 720p at 24 frames per second
- Audio: Native stereo audio, including ambient sound, dialogue and music
- Price: About $0.10 per second of streamed output, or about $6.05 per minute
- Session limit: 120 seconds maximum per session
- Access: Paid-tier Gemini API key required; no free tier
- Serving regions: United States, Western Europe and Japan
What Is Google Genie 4?
Google Genie 4 is a world model. That means it does not just generate a picture or a video. It generates a continuous, explorable world that responds to what you do inside it.
You start with a text prompt, a reference image, or both. Genie 4 then creates a moving scene in real time. You can steer through it with keyboard controls, and you can change the world while you are inside it by typing a new prompt.
The model was first announced as a research project in late October 2026. It builds on the Gemini family of models, with Gemini Omni anchoring the realistic appearance of its worlds and other Gemini models handling the reasoning that keeps simulations coherent.
The 18 November Rollout to Ultra Subscribers
Google Genie 4 first reached users through Project Genie, an experimental browser-based world builder. On 18 November 2026, Google updated Project Genie and gave access to eligible Google AI Ultra subscribers, its most expensive consumer plan.
Google AI Pro members were put on a waitlist at the same time. Google has since started sending out waitlist access to Pro users, so the subscriber rollout is widening.
The 19 November Developer Release
The bigger news for builders came a day later. On 19 November 2026, Google made Genie 4 available to developers through the Gemini API and Google AI Studio.
The model is exposed as models/genie-4 on the v1beta Interactions API. Developers create an interaction session, connect over WebRTC for browser apps or WebSocket for server code, and send a world_setup event with a text prompt, a reference image, or both.
Access requires a paid-tier Gemini API key. There is no free tier, and the service is not available through the Batch API, Flex inference or Priority inference.
How Much Does Google Genie 4 Cost?
Google Genie 4 is priced in tokens, the same unit Google uses for its language models. But the meter runs on video frames rather than words.
The price list sets $1.50 per million tokens for text or reference-image input, $9 per million for text output, and $17.50 per million for video output tokens, with stereo audio included.
Every frame costs 240 video output tokens. At 24 frames a second, that is 5,760 tokens a second. Google translates that into approximately $0.10 for each second of streamed output, or about $6.05 per minute.
Google has not yet published an exact per-token price on its main pricing page, so the $0.10 per second figure is approximate. The exact cost may shift slightly once Google publishes the final rate.
Sessions Are Capped at 120 Seconds
Each Genie 4 session can run for a maximum of 120 seconds. That limit comes from the model’s design. It renders over a rolling temporal window rather than building a permanent 3D scene, so it cannot keep a world running indefinitely.
Google’s documentation also warns that visual quality is strongest in the first 45 to 90 seconds. Visual stability may degrade after 60 to 90 seconds, as the model has to keep track of more of the world it has already generated.
What You Can Build With Genie 4
It can be used for training simulations, virtual tours, product demonstrations and creative tools. A travel company could let users explore a destination before booking. A teacher could walk students through a historical scene.
The key limitation is the 120-second session cap. Genie 4 is best suited to short, focused experiences rather than long, open-ended play.
Usage Quotas and Limits
Google sets usage quotas by paid project tier. Tier 1 accounts are limited to 2 concurrent active sessions and 50 requests per day. Tier 2 supports 5 concurrent sessions and 250 daily requests. Tier 3 permits 10 concurrent sessions and 1,000 daily requests.
Some content is blocked. Prompts showing real people, public figures or copyrighted brand intellectual property can be blocked or cut off mid-stream. Developers should plan for that when designing prompts.
What Genie 4 Cannot Do Yet
Genie 4 does not support live microphone or webcam streaming, function calling, Google Search grounding, structured outputs, or session resumption.
Frequently Asked Questions
What is Google Genie 4?
Google Genie 4 is a real-time interactive world model from Google DeepMind. It turns a text prompt or reference image into an explorable world with 720p video at 24 frames per second and native stereo audio.
When did Google Genie 4 launch?
It reached Google AI Ultra subscribers in Project Genie on 18 November 2026. Developers got access through the Gemini API and Google AI Studio on 19 November 2026.
How much does Google Genie 4 cost?
About $0.10 per second of streamed output, or about $6.05 per minute. A full 120-second session costs about $12. A paid-tier Gemini API key is required, and there is no free tier.
How long can a Genie 4 session last?
A maximum of 120 seconds. Visual stability may degrade after 60 to 90 seconds, and sessions cannot be resumed.
Who can use Google Genie 4?
Google AI Ultra subscribers can use it in Project Genie, with Pro members joining from a waitlist. Developers with a paid-tier Gemini API key can use it through the Gemini API and Google AI Studio.
Conclusion
Google Genie 4 takes world models out of the research lab and puts them in developers’ hands. At about $0.10 a second, it is not cheap, and the 120-second session cap keeps it to short experiences for now. But it is the first real-time world model in the Gemini API, and it points to a future where games, simulations and stories are generated on demand rather than built in advance. The developers who learn its limits now will be ready when those limits fall away.









