Gemini 4 Argon is out, but in the Gemini app it needs Google’s A$32.99 plan or above
Google’s new top model, Gemini 4 Argon, is rolling out to AI Pro and AI Ultra subscribers. Free and AI Plus accounts don’t get it. What it costs here.

Google has released its newest top model, Gemini 4 Argon, to paying subscribers. In a post dated 13 October, early on 14 October in Australia, the company said Argon “is rolling out today globally” in the Gemini app and AI Mode in Search for Google AI Pro and AI Ultra subscribers.
If you use Gemini for free, or pay for the cheaper AI Plus plan, nothing changes in your account. The question for everyone else is whether Argon is a reason to pay A$32.99 a month.
Who gets it
Google’s help page on Gemini models sets out which models each plan can choose in the Gemini app:
| Plan | Australian price a month | Gemini 4 Argon |
|---|---|---|
| No plan | Free | No. Most prompts use Flash-Lite |
| Google AI Plus | A$7.99 | No |
| Google AI Pro | A$32.99 | Yes |
| Google AI Ultra | A$149.99 or A$329.99 | Yes, with higher limits |
The prices are from Google’s subscriptions page as it showed for Australia on 14 October. The page doesn’t say whether they include GST, so check the total at checkout. The two Ultra prices buy different allowances: Google describes them as 5 times and 20 times the AI Pro usage limits. The same page was already advertising Argon on AI Pro and AI Ultra when we checked.
Google began moving free accounts to automatic model selection on 9 October, with most prompts going to Flash-Lite, which we covered in Google’s changes to free Gemini model access. Argon doesn’t change that. The existing Pro model stays in the model list for AI Pro and AI Ultra, alongside Argon. Google describes Argon there as offering “top-tier intelligence” and “the highest performance for complex workflows”.
For Google Search, the subscriptions page lists “Higher access to Gemini 4 Argon” for AI Pro and “Highest access” for AI Ultra. Australia is on Google’s list of countries with AI Mode. That help page doesn’t mention Argon yet, so we couldn’t confirm it is live in AI Mode here beyond the word “globally”.
What it is for
Google says Argon “excels at deep work in complex, real-world tasks across domains including coding, finance, legal, cybersecurity, and science”. The Gemini app release notes put it more simply: a model “built to take your most complex tasks from start to finish”.
For most people that means long, multi-step jobs: working through a large set of documents, building a spreadsheet model, or writing and testing code. For a quick question or a rewritten email, the faster models are still there and use less of your allowance.
Google’s launch post says Argon ranks first on the Vals Index, which it describes as measuring performance in finance, coding, legal and tax work, and reports its own internal results from using the model for coding. These are the company’s claims. We haven’t tested the model.
The limits Google doesn’t spell out
Google publishes no number of Argon prompts for any plan. Its limits page says the Gemini app has “compute-based usage limits” that refresh every five hours until you reach a weekly limit. AI Pro is described as 4 times the standard limits.
The help page on models adds that “more advanced models and higher thinking levels require more of your usage”. So the practical allowance for Argon will be smaller than for the lighter models, by an amount Google doesn’t state. If you run out, Google says subscribers can keep going with Flash-Lite.
Work accounts and developers
Google says businesses can try Argon in “Gemini for business”, through its API and in Antigravity, its coding tool. It had not published a Workspace Updates post on Argon when we checked on 14 October, so which Workspace editions include it is unconfirmed. If your team is choosing between assistants, our comparison of ChatGPT, Copilot and Gemini business plans covers how the paid tiers differ.
In Antigravity, Argon is for AI Ultra subscribers and eligible Gemini Enterprise subscriptions, according to the Antigravity announcement.
Developers pay by use. The launch post gives an introductory price of $2 per million input tokens and $10 per million output tokens, which we take to be US dollars. After the introductory period ends on 30 November 2026, that doubles to $4 and $20.
What Google says about its limits
The model card lists the usual cautions. Argon “may exhibit some of the general limitations of foundation models, such as hallucinations”, there may be “occasional slowness or timeout issues”, and its knowledge stops at March 2026, or earlier in some areas. Check figures, citations and anything recent before you rely on them.
The card also says Google “cannot rule out” that Argon has reached its internal alert levels for cybersecurity and biological information risk, and that it assesses its safeguards bring the risk below those levels. The launch post indicates the public version has cyber guardrails: it invites security defenders to apply for a version without them. Google’s safety post reports that, during training, its monitoring caught the model in rare instances trying to game tests. That matters most if you let an AI agent act on your behalf: review what it did, not only what it says it did.
Checklist
- Check whether your account can use Argon. Open the model list in the Gemini app. Google says models your plan doesn’t include show a lock icon. Argon needs Google AI Pro (A$32.99 a month) or AI Ultra.
- If you already pay for AI Pro or AI Ultra, check Argon has arrived. Google says it is rolling out, so it may not show in your model list yet.
- Check the checkout total. Google’s Australian page lists monthly prices without saying whether GST is included.
- Save Argon for long, multi-step work. Google says more advanced models use more of your allowance, and it publishes no prompt count.
- Ask your Workspace administrator before assuming a work account has it. Google hadn’t said which Workspace editions include Argon when we checked on 14 October.
- Verify figures, citations and recent events in its answers. Google’s model card lists hallucinations and a March 2026 knowledge cutoff.
Tick items as you go. Your ticks stay in this browser.




