Google DeepMind announced Genie 4, its new world model, on October 27, 2026. In DeepMind's words, "Genie 4 has been released as a research preview to trusted testers, and is coming soon to Google AI Ultra subscribers in Project Genie and early access partners in the Gemini API and Google AI Studio." Ultra subscribers did not have access as of October 27, 2026; DeepMind has opened a waitlist. Google has published no price, release date or technical specs.

#What changed?

DeepMind calls Genie 4 "our most realistic and interactive world model yet." A world model generates a scene you can move through and change in real time. DeepMind's main claims:

  • Audio and dialogue. "Genie 4 is our first real-time world model to generate audio natively." It also "introduces interactive dialogue that adapts on the fly."
  • People and physics. DeepMind says "The most dramatic leap is how Genie 4 simulates people" and that it "simulates real-world physics scenarios, such as aerodynamics, collisions, and more."
  • Steering with text. "With Genie 4, you can use natural language to modify a simulation in real-time". The model page gives examples: "Describe landscapes and events, move the camera, add new characters, or change the weather."
  • Gemini in the loop. DeepMind says "Gemini 3.8 Flash serves as a guiding agent in the loop", watching the scene, breaking goals into steps and checking incoming frames. The model "incorporates and builds on capabilities from the Gemini family", including Gemini Omni.

Users move and control the camera with the keyboard (and mouse, per the model page) and steer with text prompts. DeepMind's demos also start scenes from a reference image.

TechCrunch reported that Tejas Kulkarni, research director at Google, said during a briefing: "What we’ll increasingly see in the future are these hybrid world models, where you have a combination of Gemini writing code and text, and Genie ingesting that in real time and simulating reality at different levels of abstraction,"

#Who can use it and when?

DeepMind says it "intentionally deployed it first to trusted tester partners to get early feedback". The blog describes "cohorts of trusted testers across creative industries, journalism, education, gaming, and engineering." Google's Keyword post calls them "a limited group of trusted testers."

  • Ultra subscribers. DeepMind says Genie 4 is "coming soon" to them in Project Genie. The waitlist says "we're rolling out access gradually, starting with Google AI Ultra subscribers." It asks "Do you have a Google AI plan?" but does not promise access to other plans.
  • Developers. DeepMind says it is "coming soon" to early access partners in the Gemini API and Google AI Studio. We found no Genie entry on the Gemini API models page or changelog as of October 27, 2026, so there is no model ID or API price.
  • Regions. Google has not published Genie 4 regions. Google's plans page says Project Genie "is available in more than 140 countries" and is "Only available with a Google AI Ultra plan."
  • Age. The waitlist only asks you to confirm you are "of the minimum age required to provide data processing consent in the applicable country where you are located".
  • Timing. Google gives no date. TechCrunch wrote "In the coming weeks"; Ars Technica wrote "there’s no specific timeline yet."

#How long are sessions?

Google gives no number. The model page says: "The model can support a few minutes of continuous interaction, rather than extended sessions." TechCrunch reported Genie 4 "can maintain a consistent scene for up to two minutes, double its predecessor’s limit". The Verge reported "about two minutes long, up from one minute for the previous Genie 3 model."

For context, DeepMind described Genie 3 in August 2025 as "retaining consistency for a few minutes at a resolution of 720p" at "24 frames per second". The Project Genie app built on Genie 3 had "Limitations in generations to 60 seconds". DeepMind has not given Genie 4's resolution or frame rate.

#What DeepMind says it is for

DeepMind says Genie 4 can help "from developing training environments for robots and agents to unlocking the creation of new interactive experiences." Its examples from testers:

  • Filmmakers "can step into a live simulation of a scene and run multiple "takes" in real-time".
  • Game developers are going "from an initial idea to an explorable prototype in seconds".
  • In a research demo, "Gemini 4 Argon and Genie 4 work together to simulate Apptronik's Apollo humanoid playing a game of checkers".

#Limits and safety

DeepMind says "it remains an early, experimental model." The model page lists limits:

  • "Genie 4 is currently unable to simulate real-world locations with perfect accuracy."
  • "Clear and legible text is often only generated when it's in the input world description."
  • "Real-time audio accuracy remains an area of ongoing research and improvement."

DeepMind says it ran "internal rigorous safety evaluations" before the preview and that "all video and audio generated by Genie 4 will include SynthID".

#How does it compare?

Genie 3 launched in August 2025 "as a limited research preview, providing early access to a small cohort of academics and creators." Genie 4 names a wider path: Ultra subscribers and API partners, though neither had access as of October 27, 2026. Another lab's preview, Anthropic's Claude Duet, shows the same staged approach. For other Google releases, see Gemini for Government and Google's Agent Compute Engine.

Our analysis: The two-minute figure comes from outlets, not from Google's pages, which say "a few minutes" for both Genie 3 and Genie 4. Until Google publishes specs, readers cannot compare the two models on session length, resolution or frame rate.

#What is still unclear?

  • Price and date for Ultra or the API.
  • Specs: resolution, frame rate, latency and exact session length.
  • Regions and plans for Genie 4, including whether Ultra for Business gets it.
  • How API partners are chosen. We found no partner signup link.