Google’s Gemini 4 Argon rollout comes with an important answer for paying users: according to the Gemini app’s October 13 release notes, the new model is rolling out globally to Google AI Pro and Ultra subscribers.
If you have been waiting to try Google’s newest frontier model, that is more useful than another benchmark headline. But there is a distinction worth checking before you upgrade: access in the Gemini app, availability through the API, and access to connected tools are not the same thing. Google’s Gemini app release notes
Already paying for Pro? Check before upgrading
The dated app announcement explicitly includes Pro—not only Ultra.
It describes a rollout, however, rather than a guarantee that every eligible account will see Argon immediately. If it has not appeared for you, its absence alone does not establish that you need a more expensive plan.
There is also an inconsistency in Google’s documentation at the time of writing. The API’s latest-model guide says Argon is not available on Google AI Pro, while the October 13 app release notes explicitly announce Pro access.
For the consumer app, the dated, product-specific announcement is the stronger guide to this rollout. For API usage, check the developer documentation and your project’s billing separately. Do not treat an app subscription as proof of API entitlement. Gemini API model guide
What would make this release worth trying?
Google positions Argon around complex work: assembling presentations from scattered sources, analyzing difficult reports, and creating interactive models. These are product claims, not independent test results. Google’s release notes
The interesting test is whether it can carry a messy task to a usable result.
Consider a product launch assembled from meeting notes, a spreadsheet and an outdated brief. Writing a polished summary is the easy part. Recognizing that the spreadsheet contradicts the brief—and asking which source should win—is where the assistant starts earning its place.
That is an illustrative evaluation scenario, not a result from testing Argon.
If you get access, bring a task your current assistant handled badly. Keep the files and requirements the same. Look at what you had to fix afterward, not just how impressive the first response looked.
A new model does not need to win every benchmark to be useful. It needs to improve a task you actually repeat.
Developers get an introductory price—not a permanent one
Google’s API guide lists these rates per million tokens:
| Token type | Through November 30, 2026 | From December 1, 2026 |
|---|---|---|
| Input | $2 | $4 |
| Output, including thinking | $10 | $20 |
The listed model ID is gemini-4-argon. The guide also describes implicit caching discounts and configurable thinking levels. Google’s API pricing and model guide
For a simple illustration, 100,000 uncached input tokens plus 20,000 billable output tokens would cost $0.40 at the introductory rates and $0.80 afterward.
That is arithmetic, not a measured workload or an all-in service bill. Tool costs, repeated attempts and actual token consumption can change the total.
The practical implication: evaluate the model at the price you will pay after the launch period, too. An experiment that looks affordable in October should not become a surprise in December.
The other launch detail Mac users should notice
Alongside Argon, Google announced Antigravity integration in the Gemini macOS app. Its described workflow lets users start a coding agent, approve a plan and return to chat while the agent builds and tests in the background. Eligibility and regional conditions apply. Google’s October 13 update
That is a separate product change, not evidence that every Argon conversation automatically has coding-agent capabilities.
Still, it gives this rollout a more interesting shape: a stronger model on one side, and a shorter path from conversation to working software on the other.
For an existing Pro subscriber, the first move is straightforward: check whether Argon is available before buying anything else. Then give it one unfinished piece of real work.
A launch can get your attention. Finishing that work is what earns the next subscription payment.
Checked October 14, 2026. This article is based on Google’s public documentation, not hands-on benchmarking. Access, rollout timing and pricing may change.