Who Can Use Gemini 4 Argon? Every Google Plan Checked

Gemini 4 Argon became generally available on October 13, 2026. Whether you can use it depends on which of Google's plans you pay for, and the answer is not the same across the Gemini app, Google Workspace and the API. Google AI Pro and Ultra subscribers have it. Free users and Google AI Plus subscribers do not. A Workspace account on any core edition shows N/A unless an add-on is attached, and a developer needs a paid API tier. Those are the headlines; the rest of this post is the row-by-row check, as of October 15.

We wrote this because the launch language and the plan tables say different things. Koray Kavukcuoglu's availability post lists developers, enterprises, scientists and consumers. The Workspace help page lists five groups of editions and gives Argon a daily prompt count on only two of them. Both are true, and a business owner deciding whether to upgrade a plan needs the second one. The same question for ChatGPT is in our ChatGPT plan comparison; this is the Google version, for one model.

A note on what Google calls it. The consumer plan pages say "Gemini 4 Argon". The Workspace help page's availability table has a row labelled simply "Argon". We treat them as the same model, because no other Argon exists in Google's line-up, but the help page does not spell that out. What the model does differently from Gemini 3.1 Pro, and what changes for developers who switch, is covered in our migration checklist.

There are four ways a person or company reaches a Google model: a personal Google AI plan, a Google Workspace edition, a Workspace AI add-on, or a Gemini API key. Argon is behind each door on different terms. The grid below is the version to screenshot; the sections after it give the source for each cell and the places Google's own pages disagree with each other.

Google sells three personal AI plans above the free account: Google AI Plus, Google AI Pro and Google AI Ultra, with Ultra split into a 5x and a 20x tier. The Gemini app post of October 13 is one sentence on eligibility: Google AI Pro and Ultra subscribers can now access Gemini 4 Argon. The plans page says the same in two places. Under Pro it lists access to Argon for complex workflows and expanded access to Argon in the Gemini app and AI Mode in Search; under Ultra the same lines read "higher access". The Plus tier's model line is Gemini 3.6 Flash and Deep Research. Argon's own panel on that page adds that it is available to Pro and Ultra subscribers in select countries.

What Pro and Ultra buyers do not get is a number. Google's help page on Gemini app limits describes them as compute-based: how much you can do depends on prompt complexity, the models and features used, and chat length. The limit refreshes every five hours until a weekly cap is reached, and the page does not state the cap. The plan multipliers are the only quantities Google gives.

The example that trips people: a family on Google AI Plus, bought for storage and the 2x limit, reads that Argon is "available to consumers" and cannot find it in the model picker. It is not a bug. Plus is the one paid tier without it. The failure case on the other side is a Pro subscriber who assumes 4x means four times a known Argon quota; there is no known quota, and when the limit is reached the app drops to a Flash-Lite model rather than stopping. The help page also says some compute-heavy features may be unavailable to users without a plan during periods of high demand.

We have not printed the monthly prices for the three personal plans. Google's plans page renders them by region and they did not appear in the text we could fetch, so quoting them would mean quoting a figure we did not read on a Google page. Check the page for your country.

Two surfaces outside the chat app carry their own rule. Google Antigravity, the agentic coding environment, is listed for Google AI Ultra subscribers in Kavukcuoglu's post, and the plans page repeats it: higher access to Argon in Google Antigravity sits under Ultra, not Pro. AI Mode in Search is the opposite case: both Pro and Ultra get Argon there, according to the same post and the plans page. A Pro subscriber therefore has Argon in Search and in the app but not in the standalone coding tool.

The Gemini app post complicates this with one line: Antigravity's agentic coding interface is now available directly in the Gemini macOS app. That post is addressed to Pro and Ultra subscribers together and does not say which of them gets the embedded interface. We read it as the interface, not the full Antigravity product and its Ultra allocation, but Google has not drawn the line in writing. If you are a Pro subscriber buying the plan for agentic coding, confirm in the app before you commit.

Google's help page for Gemini Apps with a work or school account carries a table with five column groups. The first is editions where Gemini is an additional service: Business Base, Essentials Starter, legacy G Suite, Workspace Individual and a few others. The second is core-service editions at the low end: Business Starter, Essentials, Enterprise Essentials, Frontline and Nonprofits. The third is the main business tier: Business Standard and Plus, Enterprise Standard and Plus, and the legacy Gemini Business and Enterprise SKUs. The fourth and fifth are any core edition with AI Expanded Access or AI Ultra Access attached.

The Argon row reads N/A, N/A, N/A, up to 200 prompts per day, up to 500 prompts per day. That is the whole finding. A Business Plus customer paying the top small-business rate has Thinking at up to 300 prompts a day and the Pro model at up to 25 prompts per four hours, and no Argon. The add-on is the only path.

Read the Pro column alongside the Argon one and a pattern shows. On the two add-on tiers the Pro row reads N/A, which suggests the add-ons replace the Pro model with Argon rather than adding to it. Google does not explain the N/A, so we record it rather than interpret it. The context window row on the same page lists 32 thousand and 1 million tokens without a clear column mapping in the text we fetched, so we have left it out.

Education editions are not on this table. The page links to a separate article for school accounts, which we did not fetch, so this post makes no claim about Education Fundamentals, Standard or Plus. If you administer a school domain, that article is the one to read.

Google's Workspace pricing page, fetched in US dollars on October 15, lists one AI add-on under "Add-ons sold separately": AI Expanded Access, described as higher access to advanced AI capabilities in Workspace, including more Workspace Studio automations, image and video generation, and real-time speech translation. The description does not mention Argon and the page gives no price; it links to a help article. AI Ultra Access, the tier the help table pairs with 500 Argon prompts a day, does not appear on the pricing page at all.

So a buyer sizing the upgrade has the edition prices and the Argon counts, but not the add-on prices, from Google's public pages. We are not filling the gap from a reseller. The edition prices are below; the add-on cost is a conversation with Google or a partner.

Two caveats on those edition prices. They are the annual-commitment list prices on the US-dollar page; the same page shows an introductory discount for new customers whose rates changed between our fetches on October 15, limited to the first 20 users for 12 months. Enterprise is "contact sales". And the page is clear that these editions include Gemini in Gmail, Docs and the other apps; what they do not include, per the help table, is Argon in the Gemini app.

A worked example. A 40-person agency on Business Standard pays $560 a month at list for the edition. Its staff have the Thinking model at up to 300 prompts a day and the Pro model at 25 per four hours. If three people need Argon for complex drafting, the cheapest path Google's pages describe is the add-on for those seats, at a price the pricing page does not show. The alternative is personal Google AI Pro subscriptions, which carry Argon but sit outside the organisation's data controls. The help page is explicit that core-edition chats are not reviewed by humans or used to improve models; a personal plan is governed by consumer terms. That is the real trade, and it is not a price trade.

The API rule is one sentence on Google's Gemini API pricing page: Gemini 4 Argon is not available on the Free tier or on the Google AI Pro plan. The second clause matters because Google AI Pro includes AI Studio with expanded limits for Gemini 3.1 Pro and Nano Banana Pro, so a Pro subscriber reasonably expects the newest model there too. They get it in the app, not in AI Studio, where Google's AI plans page lists Argon for Google AI Ultra only. For API calls, a paid-tier key with billing attached is the requirement, at $2 per million input tokens and $10 per million output until November 30, 2026, then $4 and $20.

Two further routes exist. The Contributor tier is an invite-only private preview at half the standard rate, with the pricing page noting that data on that tier is used to improve Google's products. And a cyber-permissive variant without the public model's cyber guardrails is reserved for the Fairwind Program, which Kavukcuoglu's post says expands from October 13 to hundreds of additional partners on a rolling basis.

Google Cloud's October 13 post adds Gemini for business on web, desktop and mobile, and the Gemini platform for enterprise API access, all under the same model ID. It does not use the words "generally available"; the API changelog does. For the price tiers, caching rules and the December increase, our September 30 Argon record and the migration checklist linked above carry the detail; this post stops at who can get a key.

Put the four doors together and the excluded groups are specific. Free Google accounts. Google AI Plus subscribers. Every Workspace edition without an add-on, including Enterprise Plus. Education editions, on which Google's main table is silent. Free-tier API keys. And developers in countries outside Google's "select countries" list for the consumer plans, which we did not enumerate here. The router below is the decision as we would make it for a client, with the source for each route.

One last comparison that readers ask for: how the model itself stacks up once you have it. That is in our three-way comparison with GPT-6.1 Sol and Claude Opus 5.5. And if your organisation is weighing a Workspace add-on against a different model vendor for a specific workflow, that assessment is part of what our AI transformation team does, including the data-handling side of the personal-plan shortcut.