Demis Hassabis introduces Genie 4 with real-time physics, reasoning, and audio

What happened

Demis Hassabis shares Genie 4, describing it as an interactive world model bringing together real-time physics, reasoning, and audio, with simulations within simulations. In the quoted announcement, Google DeepMind describes exploring dynamic environments in real time, including how scenes respond and sound and how people move, react, and speak. https://twitter.com/GoogleDeepMind/status/2115114028338962672

Read next

• More changes in today’s briefingMistral releases improved Mistral Large 4 open weights for coding and cybersecurity

Reported by 10 sources

The story so far

Google DeepMind has released Genie 4, a model that generates interactive virtual worlds, as a research preview for trusted testers, with consumer and developer access still planned. Google says access is coming soon to Google AI Ultra subscribers through Project Genie and early access partners through the Gemini API and Google AI Studio; TechCrunch reports that rollout is planned in the coming weeks. Google has also opened a Project Genie waitlist. DeepMind says Genie 4 adds native audio and character dialogue, and lets users steer environments and soundscapes through natural-language prompts during simulations. Google AI says those controls include changing the weather, moving the camera and introducing new elements. DeepMind says Gemini 3.8 Flash guides multi-step events by observing scenes, planning actions and checking progress. The Verge reports that experiences can last about two minutes, up from one minute with Genie 3, though scene details and object consistency can shift over time. TechCrunch also reports problems with audio accuracy and text legibility. DeepMind says object permanence, physical interactions and long-term consistency remain research challenges.