Field intelligence for AI-first professionalsVol. II · Nº 127 · Saturday, October 31, 2026
← Phantom Notes
Nº 113AI ModelsOctober 27, 20263 min read

DeepMind's Genie 4 Is a Research Preview for Trusted Testers. AI Ultra and the Gemini API Are Coming Soon, and There Is No Genie 4 Price.

Google DeepMind announced Genie 4 on October 27 as its next interactive world model, with native audio, SynthID watermarks, and Gemini models in the loop. Access today is a research preview for trusted testers behind a waitlist; Google AI Ultra subscribers and Gemini API partners are promised access with no date. DeepMind's own limitations section says continuous interaction lasts a few minutes and that object permanence can drift. AI Ultra still starts at $99.99 a month and its Project Genie page still names Genie 3. No resolution, frame rate, or usage price is published for Genie 4. We read the announcement, the model page, and the subscriptions page.

By T.W. Ghost


Google DeepMind announced Genie 4 today, October 27, 2026: the next generation of its interactive world model, with more photorealism, more fluid movement, and native audio generated by the model itself. Gemini Omni anchors how things look, and other Gemini models, including Gemini 4 Argon in the robotics demos, help plan what happens inside the simulation. Every frame and every sound carries a SynthID watermark.

What you can do with it today is join a waitlist. Genie 4 is a research preview for trusted testers. Google AI Ultra subscribers are promised access through Project Genie, and early partners through the Gemini API and AI Studio, both "coming soon," with no date. There is no Genie 4 price, no resolution, and no frame rate on any Genie 4 page. AI Ultra still starts at $99.99 a month, and the subscriptions page still describes Project Genie as running Genie 3. The most useful section DeepMind published is its own list of limitations, and we lead with it. We read the announcement, the model page, and the subscriptions page. Here is what held up.


What Is Established

Genie 4
AnnouncedOctober 27, 2026, as a research preview for trusted testers
AccessWaitlist now; AI Ultra via Project Genie and Gemini API partners "coming soon," no date
AudioNative, generated by the model; all demo sound claimed as Genie 4 output (vendor)
Continuous interaction"A few minutes," not extended sessions (DeepMind limitations)
DriftObject permanence and visual detail can drift over longer runs or complex edits (DeepMind limitations)
WatermarkSynthID on all generated video and audio
Models in the loopGemini Omni for appearance; other Gemini models for planning; Argon in robotics demos (vendor)
Price, resolution, frame rateNone published for Genie 4
AI UltraStarts at $99.99 a month; Project Genie copy still names Genie 3

Do not inherit Genie 3's specifications. Genie 3 was described with a resolution and a frame rate. Genie 4's pages do not publish either, and we do not carry the old numbers forward.

A few minutes, with drift, in DeepMind's words. The limitations section is the vendor telling you where the demo ends. Sessions are minutes, not hours, and the thing that was behind you may not be there when you turn around.

The plan price is not a usage price. $99.99 is what AI Ultra costs; Project Genie on that plan still says Genie 3. There is no Genie 4 meter anywhere.


Verified vs Unconfirmed: The Scorecard

ClaimVerdict
Announced Oct 27 as a trusted-tester research previewVerified (DeepMind blog)
AI Ultra and Gemini API access "coming soon" with no dateVerified (blog and model page)
Few minutes of continuous interaction; object-permanence driftVerified (model page limitations)
SynthID on all outputVerified (blog)
Native audio quality; Gemini Omni and Argon in the loopVendor (demos and blog)
Genie 4 price, resolution, frame rateNone found
AI Ultra $99.99 a month; Project Genie lists Genie 3Verified (subscriptions page)
Independent hands-on reviewNone

Who Should Use It

Join the waitlist if you want Project Genie access when it opens; that is the entire available action.

Skip any plan that assumes Genie 4 today, and skip quoting a resolution or frame rate until DeepMind prints one.

The frontier is elsewhere: for video you can generate this week, the open-weight story is Pika H3 Flex and the hosted one is still Seedance 2.5. Our model comparison has the current standings on the language models side by side.


Sources


*Not sure whether a world model belongs in your pipeline before it has a price? Take the free 2-minute quiz and get matched. Then read about the video model that shipped weights the same week: Pika H3 Flex.*

Which model should you be using?

Three minutes, twelve questions, one defensible answer.

Take the quiz →