In August, we previewed Genie 3, a general-purpose world model capable of generating diverse, interactive environments. Even in this early form, honest explorers were able to create an impressive range of worlds and experiences, and found new ways to use them. The next step is to expand access with a dedicated, interactive prototype focused on creating an immersive world.
Starting today, we're rolling out access to Project Genie to Google AI Ultra subscribers in the US (18+). This experimental research example allows users to create, test and remix their own interactive worlds.
How we develop global models
A world model simulates local dynamics, predicting how they evolve and how actions affect them. Although Google DeepMind has a history of agents for specific areas such as Chess or Go, building AGI requires systems that navigate a variety of real-world environments.
To meet this challenge and support our AGI mission, we developed Genie 3. Unlike the experience that can be experienced in static 3D snapshots, Genie 3 generates a path that comes in real time as you move and interact with the world. It simulates the physics and interactions of dynamic worlds, while its successful consistency allows for the simulation of any real-world situation – from robotics and animation to modeling and fiction, to exploring locations and historical settings.
Building on our model research with trusted testers from all industries and backgrounds, we're taking the next step with a pilot study example: Project Genie.
How Project Genie works
Protocol Genie is a prototype web application powered by Genie 3, Nano Banana Pro and Gemini, which allows users to explore the immersive experience of our world model firsthand. The experience focuses on three key strengths:
