Genie 4: A new generation of world simulation

Google DeepMind announced Genie 4 on October 27, 2026, describing it as a new world model that generates interactive, photorealistic simulations in real time. The post, credited to Raia Hadsell, Diego Rivas and Hang Qi, says it is the lab's first real-time world model with native audio, producing ambient sound and dialogue that responds to users. People navigate with keyboard controls and can use text prompts to change weather, move the camera or add characters and objects as a scene unfolds.

DeepMind says Gemini models supply much of the reasoning. Gemini Omni grounds appearance and behavior, while Gemini 3.8 Flash acts as an in-the-loop agent that splits goals into steps and checks incoming frames for progress. In a robotics experiment, Gemini 4 Argon and Genie 4 simulated Apptronik's Apollo humanoid playing checkers. The company frames this pairing as an early move toward hybrid world models and calls simulation crucial to building AGI, including as training ground for robots and agents.

Access is limited for now to a research preview for trusted testers, who span film, journalism, education, gaming and engineering. Google AI Ultra subscribers in Project Genie and early access partners on the Gemini API and AI Studio are next, with a waitlist open but no date given.

DeepMind acknowledges the model is experimental: details and object permanence can drift over longer sessions or after repeated edits, and complex physics, audio accuracy, legible text and spatial accuracy remain unsolved. All generated video and audio will carry SynthID watermarks.