Genie 4 cost: Ultra price, waitlist and API access

The price question has two separate answers

There is a price for a Google AI subscription, and there will be conditions for using Genie 4. On 28 October 2026 only the first has a public figure on the US Google One page: Ultra starts at $99.99 monthly; another tier is listed at $199.99. DeepMind’s Genie 4 announcement does not identify a qualifying tier, a session allowance or a per-generation charge.

That gap matters if you are budgeting for a small studio. A subscription line in a spreadsheet cannot yet be converted into a cost per usable previs shot. No public Genie 4 API tariff is given either, so token prices for Gemini text models are irrelevant to this calculation.

What “coming soon” means for an Ultra subscriber

Genie 4 is currently a research preview for trusted testers. Google says Project Genie will bring it to Ultra subscribers, and that early access partners will get a route through the Gemini API and Google AI Studio. It provides no date or country list in this announcement. An existing Ultra account is not proof that the model has appeared in that account.

The original Project Genie launch on 29 January 2026 concerned Genie 3, initially for US Ultra users aged 18 or over. Its 60-second generation limit belongs to that prototype. It is not a confirmed Genie 4 quota. Keep the model version, the application name and your account eligibility as three separate checks.

A purchasing checklist for film and game teams

First establish whether your account has an actual Genie 4 entry point. Then check any terms displayed for sessions, exports and intended use. Save the date and the wording. Only once these are visible can a producer decide whether to charge a client for a repeatable workflow. We have no access and have not measured speed, usable takes or consistency.

For API planning, ask for the model identifier, billing unit, duration limits and access conditions. A Gemini API key alone is not evidence of Genie 4 entitlement. Do not build a live product around an endpoint name guessed from another Google model.

What you can prepare without paying

Write a short scene specification with subject, location, camera behaviour, a single action and sound. For an independent UK short, an invented rehearsal room is a safer planning example than an exact recreation of a real venue: DeepMind explicitly warns about location accuracy. Mark the specification as untested and keep a storyboard version that does not depend on the model.

Set acceptance criteria before access arrives: does the doorway remain in place after a camera turn, does the same person keep the same clothing, and does sound correspond to the visible action? These are proposed checks, not results. The official material shows potential, while the limitations describe why a polished demo cannot stand in for a production trial.

Where to register interest

Use the waitlist linked by DeepMind. A registration is not an access guarantee. Google has not published the broader Genie 4 release date, API price or country schedule. The practical budget entry today is therefore “access pending”, alongside the separately verified subscription price.

For a feature-by-feature overview and Google’s own demonstration, use the Genie 4 hub below. It includes the distinction between native sound, agent-directed actions and the continuity limits that affect longer sessions.

Related guide

Genie 4: world model, access and what Google has shown

Sources

Official documentation and demonstrations; checked on 28 October 2026.

• Google DeepMind: Genie 4 - 2026-10-28
• Genie: capabilities, limitations - 2026-10-28
• Google: introducing Genie 4 - 2026-10-28
• Project Genie (2026-01-29) - 2026-10-28
• Project Genie: prompt guide - 2026-10-28
• Genie 3 (2025) - 2026-10-28
• Google One - Google AI Ultra - 2026-10-28