Gemini 4 Argon: now available to Pro users, with a new task mode

What's new

Starting October 13, 2026, Google is rolling out Argon, a model in the Gemini 4 family, in the Gemini app. It was previously limited to Ultra subscribers, but customers with Google AI Pro now get access too. According to Google's release notes, the rollout is global across web, mobile and the Gemini desktop app for Windows and macOS. The sources do not say exactly when each individual account will get Argon, or whether special conditions apply in Denmark and the EU.

The app also gets a new toggle in the main chat that switches between regular chat and tasks (previously called Spark). At the same time, Gems in the app have been replaced by Skills, and existing Gems are moved over automatically.

What's smart

Google describes Argon as a model built to carry complex tasks "from start to finish." Examples include financial analysis and interactive 3D modeling. The idea behind task mode is that you can hand over a larger assignment instead of asking one question at a time.

On Mac, the Gemini app also gets a coding interface from Antigravity, so you can switch between chat and coding in the same app. Google is also offering 10,000 researchers free access to the highest usage tier in Antigravity for six months. Applications are open.

Cheaper or better?

Google AI Ultra costs about 100 dollars a month and provides "highest access" to Google's best Gemini models. The price of Google AI Pro is not stated in the sources. The same goes for usage limits on the individual subscriptions.

Using Argon in Antigravity (desktop app and CLI), AI Studio and Android Studio (Canary) is reserved for Ultra, according to 9to5Google.

For developers using the Gemini API, Argon costs 2 dollars per million input tokens and 10 dollars per million output tokens through November 30, 2026. Tokens are the chunks of text the model reads and writes. After that, the price rises to 4 dollars for input and 20 dollars for output per million tokens, a doubling.

How much better Argon is than competitors' models cannot be determined from the sources. There are no verified benchmark figures, and any figures from Google are self-reported and cannot simply be compared with others'.

What it's good for

Argon is aimed at longer, multistep tasks: a financial analysis, an interactive 3D model or a coding project where you want the model to work its way all the way through. For regular users with a Pro subscription, the news is that they can now get the latest model directly in the app without upgrading to Ultra. Developers and the most demanding users, however, still need Ultra to use Argon in Google's developer tools.

One caveat: Apps that can be connected to Gemini are in many cases limited to the US, select English-speaking regions, Brazil and India. That applies to the connected apps, not to Argon itself.