Google DeepMind unveils Genie 4 interactive world model

Google DeepMind has presented Genie 4, a world model that generates interactive, explorable environments from text descriptions. TechCrunch reported the launch on 27 October 2026. DeepMind calls it its most realistic and interactive world model so far.

What DeepMind says it does

According to DeepMind, users describe a landscape or event, move the camera, add characters or change the weather, and the world updates live. The company says Genie 4 builds on the Gemini family: Gemini Omni anchors simulations in real-world appearance and behaviour, while other Gemini models supply the reasoning needed to keep a simulation coherent. DeepMind also says every video and audio clip Genie 4 produces carries a SynthID watermark.

TechCrunch adds that the model generates real-time audio to match the world, including music and ambient sound.

Availability and limits

Genie 4 is a research preview for trusted testers. DeepMind says wider access will come soon, starting with Google AI Ultra subscribers in Project Genie and early access partners of the Gemini API and Google AI Studio. No date or pricing is given on the page.

DeepMind lists its own limitations. Visual details and object permanence can drift, especially in longer runs or after many prompt edits. Sessions last a few minutes rather than extended periods. Legible text usually appears only when the prompt contains it, and real-time audio accuracy is still being researched. The demonstration clips are shortened and shown for illustration, the page says.

Why it matters

This is our analysis. DeepMind frames world models as a way to train and test AI agents in realistic environments, which makes Genie 4 a research tool first. With access limited to trusted testers and no benchmarks published on the page, how useful it is outside demos remains untested.