Google DeepMind unveils Genie 4, a real-time multimodal world model
Image: Google DeepMind / YouTube · video-thumbnail
Genie 4 generates interactive, steerable simulations with sound, and DeepMind pairs it with Gemini 4 Argon to simulate a humanoid; no physical-robot results are reported.
Google DeepMind introduced Genie 4 on 27 October 2026, a world model that generates interactive simulations in real time. It is natively multimodal: audio, including music, ambient sound and dialogue, is generated alongside the visuals. Users steer a scene with keyboard controls and text prompts.
DeepMind's post also shows Genie 4 working with Gemini 4 Argon to simulate Apptronik's Apollo humanoid playing checkers. Argon breaks the task into steps and passes them to Genie 4 as text captions during rendering. DeepMind lists training environments for robots as a potential use.
The post reports no transfer to a physical robot. It describes simulation as a way to test how AI systems handle environments before real-world deployment, which is a stated goal rather than a demonstrated result.
Genie 4 is a research preview for trusted testers. DeepMind says it is coming soon to Google AI Ultra subscribers and to early-access partners in the Gemini API and Google AI Studio.
Sources
- Google DeepMind Primary source
