TellWorldAI World Archive

Home›Models

Model

Genie 4

Genie 4 is Google DeepMind's world model, released on 18 November 2026 to Google AI Ultra subscribers in Project Genie and to developers in the Gemini API the next day. It generates interactive worlds with native audio.

Status AvailableModality Image, Audio, VideoOpen weights No Era Generative AI for everyone

License
Proprietary
Access
API, App
Release date
Developer
Google DeepMind

Background

Genie 4 is a world model from Google DeepMind. A world model generates a virtual environment that a user can move through, and that environment keeps changing in response to input. Google rolled Genie 4 out on 18 November 2026 and called it its most realistic and interactive world model yet. It follows Genie 3, which Google also updated on the same day.[1]

What it does

Users can prompt mid-session to change a generated world, for example by shifting the weather or adding a character, and the edits take effect in real time.[1] The visuals come with real-time native audio that adapts as the scene unfolds.[1]

In Project Genie, an agent builds worlds from a natural-language conversation by orchestrating Google's multimodal models.[1] Every Genie 4 output includes imperceptible SynthID watermarking.[2]

Access

On 18 November 2026 Genie 4 rolled out to eligible Google AI Ultra subscribers in a new Project Genie experience. Google AI Pro members can join a waitlist for Project Genie.[1]

On 19 November 2026 Genie 4 became available to developers in the Gemini API through the Interactions API and Google AI Studio.[2] The API version generates continuous 720p worlds at 24fps with synchronized native audio. A world can start from a text prompt, a reference image, or both.[2]

Genie 3 updates

Google updated Genie 3 on the same day. Sessions now run for up to 3 minutes per generation, up from 60 seconds. Genie 3 also gained support for 360-degree panorama image grounding and for linking multiple worlds into one sequence.[1]

Position

Genie 4 now reaches two groups: subscribers who use Project Genie, and developers who build with the Gemini API.[1][2]

Terms used in this entry

A world model learns how environments change over time, so it can predict what a scene should look like after a user acts. Because the output is produced while the user interacts with it, a session can be steered step by step, which is different from watching a fixed video. A frame rate of 24fps means 24 images are shown each second, and a 720p resolution describes the number of pixels in each frame. Synchronized audio means the sound is produced along with the picture so the two stay aligned as the scene changes.

Version history

4
  1. 2024
    • Feb 23, 2024 Genie

      A Google DeepMind generative world model with 11 billion parameters that turns text, images or sketches into controllable virtual worlds, described as the first generative interactive environment trained on unlabelled internet videos without action labels.

    • Dec 4, 2024 Genie 2

      Genie 2 is a foundation world model from Google DeepMind, announced on 4 December 2024, that turns a single prompt image into a playable 3D environment that a person or an AI agent can explore with keyboard and mouse controls.

  2. 2025
    • Aug 5, 2025 Genie 3

      Genie 3 is Google DeepMind's general-purpose world model, announced on 5 August 2025. It turns text prompts into interactive 720p environments that can be navigated in real time at 24 frames per second, and was released as a limited research preview.

  3. 2026
    • Nov 18, 2026 Genie 4 · This page

      Genie 4 is Google DeepMind's world model, released on 18 November 2026 to Google AI Ultra subscribers in Project Genie and to developers in the Gemini API the next day. It generates interactive worlds with native audio.

20261

  1. ReleaseGoogle DeepMind releases Genie 4 in Project Genie and the Gemini API

    On 18 November 2026 Google DeepMind announced Genie 4 in Project Genie, then on 19 November made it available to developers in the Gemini API and Google AI Studio. Genie 3 also got longer sessions the same day.

Developed by1

Google DeepMind

Successor of1

Genie 3

Notes and sources

  1. Genie 4 is now availablehttps://deepmind.google/blog/genie-4-is-now-available/
  2. Genie 4 now available through the Gemini APIhttps://blog.google/innovation-and-ai/technology/developers-tools/genie-4-in-gemini-api/

Cite this page: Genie 4. TellWorld, https://www.tellworld.net/models/genie-4. Updated 2026-11-20