Learn about this first
Google DeepMind has released the Genie 4 world model as a research preview for “trusted testers.” It is not yet available to general consumers or developers. Google says access will expand in the coming weeks to Google AI Ultra subscribers through Project Genie and to early-access partners through the Gemini API and Google AI Studio.
Genie 4 succeeds Genie 3 and can generate explorable virtual worlds frame by frame in real time from text or image prompts, without pre-rendering environments or programming interactive elements. Google says it offers more photorealistic visuals and more accurate physics than its predecessor, and lets users change the weather, camera, and scene elements using natural language while a simulation is running. It is also DeepMind’s first real-time world model to generate native audio, including ambient sounds, music, and character dialogue that responds to users.
According to Google and several reports, Gemini models handle reasoning and planning within the simulation. Gemini Omni anchors the world’s appearance, while Gemini 3.8 Flash interprets prompts, breaks down goals, and replans inside the simulation loop. Google says such virtual worlds could eventually be used to train robots and AI agents. These capabilities reflect Google’s descriptions at launch; real-world performance remains subject to feedback from testers.
AI-generated from reports · updated 17 days agoLatest developments
Google DeepMind releases the Genie 4 world model, which can create and control interactive simulations in real time
Popularity trend
Event progress3
Genie 4 research preview rolls out to testers
Google: Genie 4 research preview opens to trusted testersGenie 4 will be available through an API for the first time
Logan Kilpatrick: Genie 4 world model is coming to the APIGenie 4: A new generation of world simulation
Google DeepMind: Blog: Google DeepMind releases real-time interactive world model Genie 4