Google Opens Gemini 4 Argon to AI Pro, Ultra and Paid API Users

Gemini 4 Argon spent its first two weeks locked behind a cybersecurity partner program. As of October 13, anyone paying for Google AI Pro or AI Ultra can select it, and developers can call it through the paid API.

Gemini 4 Argon opens to Pro and Ultra subscribers

Google announced on October 13 that Gemini 4 Argon is now available to paying users, roughly two weeks after the model's tightly limited debut. The expansion covers Gemini app subscribers on the AI Pro and AI Ultra plans, developers building through the paid Gemini API and AI Studio, and Google AI Ultra users inside Antigravity, the company's agentic coding tool.

For anyone who watched the original announcement and wondered when they could actually try it, that's the headline. Argon was introduced on September 30 to a small group of cyber defenders through what Google calls the Fairwind program. The model existed. Most people just couldn't reach it.

Access is rolling out in stages. A paid subscription doesn't guarantee the model shows up in your account the moment you read this. Google says the rollout is gradual, so two Ultra subscribers can open the same app on the same day and see different options.

What Gemini 4 Argon is built to do

Google is positioning Argon for complex, multistep tasks rather than everyday chat. The pitch is a model that can hold a long chain of steps together: plan, execute, check its own work, and keep going. Coding, finance, legal work, cybersecurity, and scientific research are the areas the company names.

The most concrete capability claim is output length. Argon supports up to roughly one million tokens of output in a single run, up from tens of thousands on earlier Gemini models. The catch is that input, reasoning, and output share the same one-million-token window, so a long question eats into the space left for the answer.

Google's benchmarks put Argon at the top of the Vals Index, a ranking that scores models on economically relevant work such as finance, coding, law, and tax. Those numbers come from Google, and third-party leaderboards that reuse vendor data. Independent testing at this scale is still thin, and Google's own framing leans on its internal use cases.

Which Google plan you need for Gemini 4 Argon

The entry points don't line up, and that trips people up. Google AI Pro starts at about $19.99 a month and gets you Argon in the Gemini app. AI Ultra costs more and adds access in AI Studio and Antigravity.

Here's the part worth flagging. Buying Pro for the app won't get you Argon in Antigravity. Google's published support tables list Argon for Ultra and Enterprise tiers there, and exclude the older standalone Antigravity IDE. So if coding inside a repository is your reason for paying, Pro won't get you there.

The cleanest way to think about it: pick the entry point by the task, then check that specific account's permissions. Chatting with documents and asking questions is an app feature. Working across a codebase is an Antigravity feature. Calling a model thousands of times from your own software is an API feature with its own billing.

So which plan do you actually need? If you only want Argon in the Gemini app, Pro covers it. If you want to build with the model, budget for the API instead.

API pricing changes December 1

Developers pay per token, and the current rates are promotional. Through November 30, input runs $2 per million tokens, output runs $10 per million, and cached input carries a 95% discount. On December 1, those rates double: input to $4 per million and output to $20 per million.

That's steep. That's a meaningful jump for anyone building a product that leans on Argon at volume. The gap between a promotional rate and the standing rate is where a lot of side projects quietly die. If you're weighing whether Argon fits your budget, run the math against the December number, not the October one.

Google hasn't said whether the promotional period will be extended. Treat the November 30 date as fixed for planning purposes. If you want a sense of scale, a project that burns ten million output tokens a month costs about $100 at the promotional rate and about $200 after December 1. That gap decides how generous you can be with long agent runs.

What Gemini 4 Argon's wider rollout changes

Two things shift now that the model is generally reachable. First, the market gets to test Google's claims instead of reading them. Subscribers can line Argon up against the models they already pay for, and developers can measure whether it finishes a task accurately enough to justify the cost.

A benchmark lead doesn't settle that question. The everyday test is whether the model handles a half-formed request, notices when it's missing information, and closes the loop without handing you clean-up work.

Second, the safeguards get stressed. Argon's debut was deliberately narrow, with the cybersecurity capabilities held back. A much larger pool of users, including organizations running it against live systems, is a harder test than a curated partner list. Google will need to show the guardrails hold at scale, and that's a claim no announcement can settle in advance. Time will tell.

Google also says Argon connects to more than 40 apps, and the exact integrations depend on the product and the account. A model that reaches into the tools you already use is more useful, and also more sensitive. Permissions, confirmation steps, and whether you can inspect what it did all start to matter as much as raw capability.

Should you switch to Gemini 4 Argon?

That depends on what you pay for today and what you'd give up. If you're already on a Google AI plan for the Gemini app, Argon is a free upgrade at the same price, and there's little reason not to try it on a demanding task. If you're weighing it against a rival subscription, run it on real work first, because a benchmark score doesn't tell you how it handles your particular mess.

For developers, the math is more about the December price than the October one. The promotional rate makes Argon look cheap next to premium rivals. The standing rate narrows that gap, and a workload that's only viable at the lower number will stop being viable in December. Test on the number you'll actually pay.

Argon is available now to eligible subscribers as the rollout reaches their accounts, and to developers through the paid API. If your plan is Pro and you signed up for Antigravity coding, expect disappointment until you move up a tier. Watch the December 1 price change before you commit a production workload.