Genie 3

Genie 3 is Google DeepMind’s world model. Give it a prompt or an image and it generates an environment you can actually move through, keeping the scene consistent for several minutes rather than dissolving after a few seconds like most generated video.

It is often mentioned in the same breath as Sora and Marble as an example of where AI might go after chatbots. Researchers argue about whether systems like this genuinely model how the world works or just produce very convincing footage of it.