Google made its Genie 4 world model available to developers through the Gemini API and Google AI Studio on 19 November, a day after launching it in Project Genie for eligible Google AI Ultra subscribers, according to Google’s two announcement posts. Neither Google post states a price, a session limit or a model ID. Two secondary pages put the price at about $0.10 per second of streamed output.

What Google says

  • 18 November post (by Diego Rivas, Group Product Manager, Google DeepMind, and Pavan Kumar, Group Product Manager, Google Labs): Genie 4 is “our most realistic and interactive world model yet”. Project Genie is rolling out to “eligible Google AI Ultra subscribers”, and Google AI Pro members can join a waitlist. Users can “prompt mid-session to change the world you created”, with real-time native audio. The post says all video and audio generated in Project Genie includes SynthID watermarking and that all model inputs and outputs go through safety screenings, and that Genie 3 and Genie 4 “remain world models in the frontier research space”, naming camera latency, intricate physical interactions, real-time audio precision and long-term scene permanence as open challenges.
  • 19 November developer post (no byline): Genie 4 is available through the Gemini API, via the Interactions API, and in Google AI Studio. It describes “continuous 720p worlds at 24fps” with synchronized native audio, started from a text prompt, a reference image or both, with real-time keyboard controls alongside live text prompts. It says every output includes imperceptible SynthID watermarking and names prototyping interactive experiences, visualizing designs for spatial concepts and training robotics and AI agents as uses. On price it says only “Review pricing details” and links to the AI Studio documentation.

What the secondary pages add

PPC Land (Luis Rijo, 19 November) and TestingCatalog (19 November) report the model ID as models/genie-4, sessions capped at 120 seconds, a paid-tier Gemini API key, and availability in the US, Western Europe and Japan. Both put the price at about $0.10 per second. TestingCatalog gives $1.50 per million input tokens, $9 per million text output tokens and $17.50 per million video output tokens; PPC Land says the exact rate is not public and describes metering as 240 video output tokens per frame, or 5,760 per second at 24 frames per second. 5,760 tokens at $17.50 per million is about $0.10 per second, so a 120-second session would cost about $12 (Lens arithmetic). PPC Land also says visual stability may degrade after 60 to 90 seconds and that landmarks, doors and buttons may shift if the user looks away for more than 8 seconds.

Where reports differ

  • Price: Google’s posts give none. TestingCatalog gives token rates; PPC Land says the rate is not public. A “$6 per minute” figure in other headlines is arithmetic on the per-second price, not a Google price.
  • Who can use the API: Google’s developer post states no eligibility rule; the secondary pages say a paid-tier key is required. Project Genie access is limited to eligible Ultra subscribers on Google’s 18 November post.
  • Pro waitlist: PPC Land says Pro members “can join the waitlist”; a post quoted by TestingCatalog says Pro users have started receiving access from the waitlist.
  • Concurrency: The Defiant’s summary says concurrent sessions are limited by billing tier; PPC Land says no concurrency limits are published.
  • Watermarking: Google’s developer post says every output carries SynthID; PPC Land notes that Google’s Genie developer pages do not mention it.

Sources and notes

Lens read Google’s two announcement posts on blog.google, PPC Land and TestingCatalog. The Defiant’s page returned only its headline and summary. The AI Studio documentation and Google’s Gemini API pricing page were not opened (a fetch of the documentation page was denied), so the model ID, session cap, regions and price rest on the secondary pages. The TestingCatalog article says it was drafted with AI assistance and reviewed by an editor.

Image: File photo of a Google building at the Googleplex in Mountain View, California, 2016, by Asoundd, CC BY-SA 4.0, via Wikimedia Commons. It is a stand-in and is not a Google DeepMind site.