Genie 4

A world model generates an interactive environment frame by frame and reacts as you move through it, rather than just producing a static image or video clip. Genie 4 is Google DeepMind's latest version of this capability, announced on October 27, 2026. What makes it different from its predecessor Genie 3 is that it generates audio natively alongside the visuals, including ambient sound, music, and interactive dialogue from characters that respond to what's happening in the scene in real time.

The Gemini integration is the other headline feature. Gemini 4 Argon runs as a persistent director inside the simulation: it observes each generated frame, breaks goals into steps, checks that each step succeeded, and adjusts its plan when the world shifts or the user intervenes. This makes Genie 4 closer to a live, steerable environment than a video generation tool. Google also demonstrated using it for robotics training, where Gemini plans tasks that Genie simulates, letting teams test robot behaviors in a generated world before trying them in a physical one.

As of late October 2026 it is a research preview available to trusted testers only, with broader access coming to Google AI Ultra subscribers through Project Genie and to developers via the Gemini API and AI Studio. For builders, the most relevant near-term use cases are game prototyping, interactive experience design, and synthetic training environments for embodied AI.