Genie 4 Access: How to Try Google's World Model and What It Costs
Genie 4 is in limited research preview and rolling out to Google AI Ultra's $250/month plan. Here's who gets access and when.
What is Genie 4 and who can use it right now?
Genie 4 is Google DeepMind’s latest world model, a system that generates an interactive, explorable environment from a text prompt instead of a fixed video clip. Right now it’s in a research preview, meaning access is limited to a small group of testers in film, gaming, education, journalism, and engineering. Google has said it will expand access in the coming weeks to subscribers of Google AI Ultra, its top-tier consumer AI plan priced around $250 a month in the US, through a feature called Project Genie. For most people reading this, the honest answer is: not yet, unless you’re willing to pay for the most expensive plan Google sells.
TL;DR
- Genie 4 is currently restricted to a research preview, available only to a small set of trusted testers working in film, gaming, education, journalism, and engineering.
- General access will come through Google AI Ultra, Google’s premium subscription tier, priced at roughly $250 per month in the US.
- Project Genie is the delivery mechanism, the branded feature inside Google’s ecosystem that lets Ultra subscribers interact with Genie models.
- Business access is also planned, with Google rolling the technology out to some early access developers through its enterprise AI tools.
- Sessions are short right now, with earlier versions of Project Genie capping interactions at 60 seconds, reportedly due to the cost of running the model.
- Pricing is likely to drop over time, following the same pattern as every other AI product that launched expensive and later became cheap or free as hardware and efficiency improved.
- The compute cost behind each session is enormous, since every frame is generated live and individually rather than streamed from a pre-made file, which is the core reason access is gated behind a high price.
How do you actually get access to Genie 4?
There are two tracks. The first is the research preview, which is invite-only and limited to specific professional categories: filmmakers, game developers, educators, journalists, and engineers. This group got early hands-on time, which is why demo clips (a motorcycle ride along a coastline, a diver exploring a sunken cave, a neo-gothic city filled with dragons) started circulating from named users before any public launch.
The second track, and the one that matters for most people trying to use this, is Google AI Ultra. Google has said Genie 4 will roll out to Ultra subscribers through Project Genie “in the coming weeks” from the time of its announcement. Ultra is Google’s most expensive consumer AI subscription, and access to Genie 4 is being positioned as one of its headline perks rather than a separate purchase.
There’s also a quieter third path: Google is extending early access to some developers through its business-facing AI tools, suggesting a parallel API or enterprise offering may eventually exist alongside the consumer product.
Why does Genie 4 cost so much to access?
The price tag traces directly back to how the model works. A streaming video file, like something you’d watch on Netflix, is rendered once and then copied to millions of viewers. Genie can’t do that. Every person’s session is unique and changes in real time based on what they type or where they move, so the model has to generate each frame live, individually, for that one user. That works out to roughly 24 AI-generated images per second, meaning a single minute of use produces well over a thousand freshly generated frames, each one consistent with the frames before it and reactive to whatever the user just did.
That computation runs on Google’s own specialized AI chips inside Google’s data centers. Earlier versions of Project Genie reportedly capped sessions at 60 seconds specifically because of compute costs, and Google’s own researchers have acknowledged that extending session length adds significant cost for comparatively little benefit. Put simply: this is one of the most computationally expensive consumer AI products currently available, and the $250/month price point reflects that.
Is the $250/month Ultra plan worth it just for Genie 4?
That depends entirely on what you want it for. If you’re a developer, researcher, or creator who wants early hands-on experience with one of the more technically ambitious AI releases of the year, the access might be worth it to you simply for the head start. If you’re a casual user curious about the novelty, it’s a steep price for a feature that still has real limitations: scenes drift and lose consistency over longer sessions, objects can shift shape or disappear, and text rendered inside these worlds is often unreadable.
It’s also worth remembering how fast this category has moved. The first Genie model, released in early 2024, produced blurry two-dimensional worlds at roughly one frame per second, basically a slideshow. By late 2024, Genie 2 managed rough 3D worlds for 10 to 20 seconds. Genie 3 arrived in August 2025 running at 24 frames per second for about a minute. Genie 4 adds sound, sharper realism, and roughly double the session length of its predecessor. Given that trajectory, paying a premium today buys early access to something that will likely be cheaper, faster, and more capable within a year or two, not a finished, stable product.
When will Genie 4 become more widely or cheaply available?
Nobody at Google has given a specific timeline beyond the near-term Ultra rollout, but the pattern with past AI products offers a reasonable guide. Expensive, gated capabilities tend to trickle down to cheaper tiers as the underlying hardware gets faster and researchers find ways to make the models smaller and more efficient. Waymo’s own world model work is a relevant example: the company built a smaller, cheaper version of its simulation model specifically so it could run more simulations at lower cost. There’s no reason to assume Genie won’t follow the same path.
For now, expect wider exposure to Genie 4 to come indirectly, through social feeds filled with other people’s generated worlds, before it becomes something most casual users can access directly without paying for Ultra. That’s consistent with how most high-end AI launches have unfolded: creators and early testers get access first, and everyone else sees the results secondhand before the tool itself becomes affordable or free.
Frequently Asked Questions
How much does Genie 4 cost to access?
Direct access comes through Google AI Ultra, which costs around $250 a month in the US. There’s no standalone price for Genie 4 itself; it’s bundled as a feature of the Ultra subscription through Project Genie.
Can developers get access outside the Ultra plan?
Google has said some early access developers will be able to try Genie 4 through its business-facing AI tools, separate from the consumer Ultra rollout, though details on pricing or scope for that track haven’t been made public.
Why are sessions in Project Genie so short?
Earlier versions limited sessions to around 60 seconds, reportedly because of how expensive it is to compute. Every frame is generated live for each individual user rather than streamed from a pre-rendered file, which makes longer sessions costly to run at scale.
Will Genie 4 ever be free or cheaper?
There’s no announced timeline, but AI tools historically get cheaper as chips improve and models become more efficient. It’s reasonable to expect the cost per session to fall over time, the same way earlier expensive AI capabilities eventually became widely accessible.
Is Genie 4 available to the general public right now?
Not yet. It’s currently limited to a research preview for select testers, with a wider rollout planned for Google AI Ultra subscribers in the near term rather than a fully public, unrestricted release.
