Google DeepMind has revealed Genie 3, its latest foundation world model that can be used to train general-purpose AI agents, a capability that the AI lab says makes for a crucial stepping stone on the path to \u201cartificial general intelligence,\u201d or human-like intelligence.\u00a0<\/p>\n
\u201cGenie 3 is the first real-time interactive general-purpose world model,\u201d Shlomi Fruchter, a research director at DeepMind, said during a press briefing. \u201cIt goes beyond narrow world models that existed before. It\u2019s not specific to any particular environment. It can generate both photo-realistic and imaginary worlds, and everything in between.\u201d<\/p>\n
Still in research preview and not publicly available, Genie 3 builds on both its predecessor Genie 2<\/a> (which can generate new environments for agents) and DeepMind\u2019s latest video generation model Veo 3<\/a> (which is said to have a deep understanding of physics).\u00a0<\/p>\n With a simple text prompt, Genie 3 can generate multiple minutes of interactive 3D environments at 720p resolution at 24 frames per second \u2014 a significant jump from the 10 to 20 seconds Genie 2 could produce. The model also features \u201cpromptable world events,\u201d or the ability to use a prompt to change the generated world.<\/p>\n Perhaps most importantly, Genie 3\u2019s simulations stay physically consistent over time because the model can remember what it previously generated \u2014 a capability that DeepMind says its researchers didn\u2019t explicitly program into the model.\u00a0<\/p>\n Fruchter said that while Genie 3 has implications for educational experiences, gaming<\/a> or prototyping creative concepts, its real unlock will manifest in training agents for general-purpose tasks, which he said is essential to reaching AGI.\u00a0<\/p>\n \u201cWe think world models are key on the path to AGI, specifically for embodied agents, where simulating real world scenarios is particularly challenging,\u201d Jack Parker-Holder, a research scientist on DeepMind\u2019s open-endedness team, said during the briefing.<\/p>\n Techcrunch event<\/p>\n \n\t\t\t\t\t\t\t\t\tSan Francisco<\/span> Genie 3 is supposedly designed to solve that bottleneck. Like Veo, it doesn\u2019t rely on a hard-coded physics engine; instead, DeepMind says, the model teaches itself how the world works \u2014 how objects move, fall, and interact \u2014 by remembering what it has generated and reasoning over long time horizons.\u00a0<\/p>\n \u201cThe model is auto-regressive, meaning it generates one frame at a time,\u201d Fruchter told TechCrunch in an interview. \u201cIt has to look back at what was generated before to decide what\u2019s going to happen next. That\u2019s a key part of the architecture.\u201d<\/p>\n That memory, the company says, lends to consistency in Genie 3\u2019s simulated worlds, which in turn allows it to develop a grasp of physics, similar to how humans understand that a glass teetering on the edge of a table is about to fall, or that they should duck to avoid a falling object.<\/p>\n Notably, DeepMind says the model also has the potential to push AI agents to their limits \u2014 forcing them to learn from their own experience, similar to how humans learn in the real world.<\/p>\n
\n\t\t\t\t\t\t\t\t\t\t\t\t\t|<\/span>
\n\t\t\t\t\t\t\t\t\t\t\t\t\tOctober 27-29, 2025<\/span>\n\t\t\t\t\t\t\t<\/p>\n<\/p><\/div>\n<\/p><\/div>\n<\/div>\n![\"\"](\"https:\/\/techcrunch.com\/wp-content\/uploads\/2025\/08\/Prompt-to-World.gif?w=680\")
![\"\"](\"https:\/\/techcrunch.com\/wp-content\/uploads\/2025\/08\/Prompt-Event.gif?w=480\")