Genie 3
AI Model
World Model
Current
Playable worlds from a text prompt — DeepMind's world-model bet.
You're looking at an AI model record: who built it, what it can do, and the hardware and data behind it.
Genie 3 (August 2025) generates interactive, navigable environments at 720p/24fps from text — worlds that stay consistent for minutes and respond to 'promptable events.' DeepMind frames world models as the path to robot and agent training at scale, and as a step on the road to AGI.
How this connects to the rest of the stack
This record's links at a glance — click a group to jump to it.
Developed by
Competes with
Related
Sources
Record reviewed as of 2026-01-15 — see Methodology.