Google has a whole new world for people to play with—but only for a minute. This week, the company released Project Geniewhich the company calls a “general-purpose world model” capable of creating interactive environments. First revealed to a small group of invite-only testers in August of last year, the model, known as Genie 3, is now launching to Google AI Ultra subscribers in the US, so you can get your hands on it for the low, low price of $250 per month.

The fact that Google shows a model to the world is interesting in itself. Unlike a large language model (LLM), the underlying technology that powers most consumer-facing AI tools including Google’s own Gemini, which uses large amounts of training data provided to them to predict the most likely next part of a sequence, global models are trained on real-world dynamics, including physics and spatial properties, to create a simulation of how the physical environment behaves.

World models are the AI ​​approach favored by Yann LeCun, the former chief scientist of Meta AI. LeCun believes (perhaps correctly) that LLMs will never achieve artificial general intelligence, the point where AI can match or surpass human capabilities in all domains. Instead, he believes the world’s models can chart a path toward that goal, and he recently joined a startup that’s all in that bet. This is an oversimplification, but the idea is that LLMs can only identify patterns, while world models will allow AI to run tons of simulations to understand how the world works and extrapolate new conclusions.

Google playing in this world certainly gives some legitimacy to the idea that world models offer something that LLMs cannot, and there is no denying that the preview videos that came out in the early days of Project Genie were quite impressive, albeit brief. Google prevents users from creating 60 seconds worth of their world, which the company also says “may not look completely true-to-life or always closely follow prompts or images, or real-world physics,” – which is to say, it doesn’t work. The outputs are now 720p videos delivered at 24 frames per second, in Ars Technicaand users sometimes complain that it quite laggy in practice.

I got early access to Project Genie from @GoogleDeepMind ✨

It’s unlike any other realtime world model I’ve tried – you create a scene from text or a photo, and then design a character that can explore it.

I tried several prompts. Here are the standout features 👇 pic.twitter.com/I6CPJzPzIG

– Justine Moore (@venturetwins) January 29, 2026

That’s fine for something in beta, though it speaks to the limitations of the company’s model, which suggests the world may be smaller than you think. While users are hyping the feature as if it’s about to put video game developers out of business, it’s probably worth pumping the brakes on that concern for now.

Project Genie is launching for AI Ultra members in the USA. It’s an experimental tool that allows you to create and explore endless virtual worlds, and I’ve never seen anything like it. It’s early, but it’s not true anymore.

Nano Banana Pro + Project Genie = My low-poly… pic.twitter.com/mkMEw2GxxU

– Josh Woodward (@joshwoodward) January 29, 2026

Google’s Genie 3 also takes a different approach to world models than LeCun envisioned. The model, available through Project Genie, essentially creates a continuous video-based world. Users can navigate like in a video game, but in theory, AI agents can also endlessly run through worlds to understand how things work. LeCun’s idea when he was at Meta was to create the Joint Embedding Predictive Architecture (JEPA), which embeds a model of the external world in an AI agent.

But again, the fact that Google is pushing a global model says something. Yes, the company is going through all the same issues that come from releasing other image and video processing models like OpenAI’s Sora 2, which is used to create all kinds of potential copyright infringement. Early Project Genie output is faithful reproduction of the worlds of Nintendofor example, and that will probably cause some issues. But it also suggests that even the biggest players in this AI space recognize that LLMs can hit a wall.

Holy moly… Genie 3 created this mock 3D game world from Breath of the Wild.

How I did it + prompted the comment. https://t.co/IrwlZ1pTMs pic.twitter.com/H33an42YNd

— Min Choi (@minchoi) January 29, 2026

That said, there’s a reason Google has put a hard cap on Project Genie for now. If you think it costs a lot to train and operate a text-based model, just think about what it takes to create a fully-fledged real-world simulation. It takes tons of high-dimensional data to understand everything from seeing a world to how physics works, and it takes a lot of processing power to run. So, for now, the worlds may look spacious but are kept small in practice.