Google DeepMind releases Genie 4 world model in research preview

What happened

Google DeepMind released its newest world model, Genie 4, in research preview on Tuesday, October 27, 2026 [1]. This model is designed to generate photorealistic and interactive simulated worlds. Key technical updates include:

• Scene Duration: The model can maintain a consistent scene for up to two minutes, which is double the one-minute limit of the previous Genie 3 model [1, 3].
• Audio Generation: It generates real-time audio that matches the simulation, including ambient noises, music, and interactive dialogue [1].
• Real-time Interaction: Users can type prompts during a simulation to change the scene, such as adjusting the weather, camera angles, or adding new objects and people [1].
• Integration: The model uses Gemini 3.8 Flash to assist with simulations [3]. Gemini acts as a director to plan steps and check frames while Genie creates the visuals [1].
• Physics and Characters: The model offers improved physics and more humanlike characters compared to its predecessor [1].

During a research experiment involving Gemini Robotics, a Gemini 4 Argon model was used to break down tasks into steps that Genie 4 simulated in real time [1]. Tejas Kulkarni, a research director at Google, stated that future developments may include hybrid models where Gemini writes code and text while Genie simulates reality [1].

Despite these advancements, Google DeepMind engineers noted certain limitations. Diego Rivas, a Google DeepMind engineer, mentioned that the technology is in early stages [1]. The company acknowledged that details and objects can drift during longer simulations, and they have not yet ensured accurate audio or legible in-world text [1].

Availability for Genie 4 is as follows:

• Current status: Available now as a research preview for “trusted testers” [1].
• Upcoming rollout: In the coming weeks, it will be available to Google AI Ultra subscribers through Project Genie, as well as early access partners of the Gemini API and Google AI Studio [1]. A waitlist has been opened for those wishing to try it via Project Genie [3].

Who is affected

Google identifies several potential use cases for this technology:

• Filmmakers: For use in prototyping, storyboarding, and rapid ideation [1].
• Game Developers: For prototyping game ideas [3].
• Robotics Researchers: For creating training scenarios and testing robot behaviors in simulation before real-world application [1, 3].
• Educators: For interactive learning, such as simulating historical scenes where users can interact with subjects [1].