Gemini 4 Argon reaches paid users, with introductory API prices ending in November

Google began a broader rollout of Gemini 4 Argon on 13 October, opening the model to paid Gemini API customers and Google AI Pro and Ultra subscribers through the Gemini app. Its introductory API rates expire on 30 November 2026; the announced input and output prices then double.[1]

For a reader who has followed Argon as an announced capability, this is a concrete access story. The question is no longer only what the model might do, but which product supplies it and how the bill changes when an experiment becomes a continuing workload.

Three ways into the same model

The developer release describes Argon as a model for sustained coding and agentic work, including changes that stretch across a codebase rather than a single function. Developers can use the paid Gemini API; Google also identifies AI Studio and Antigravity routes for Ultra subscribers. Google’s AI Studio subscription guide reserves Argon for Ultra rather than Pro, and direct API use is billed separately. The Pro entitlement in the Gemini app is a separate access arrangement.[2][4][3]

In the consumer app, Google is introducing a switch between chat and tasks. The company says tasks can continue while the user is away from the computer. Its separate announcement also places Antigravity’s coding interface inside the Gemini macOS app and describes connections to more than 40 apps, including OpenTable, Instacart, Adobe and Shopify. Pro and Ultra subscribers are the stated audience for Argon and tasks.[3]

That changes the shape of an assignment. “Explain how I could reorganise this project” calls for an answer. “Reorganise this project and return the changed files” calls for work that persists beyond the conversation. This is an interpretation of the product direction, not a report that every connected service supports every action.

What the introductory prices buy

For a hypothetical job using 900,000 uncached input tokens and 100,000 output tokens, those rates imply US$2.80 during the introductory period: US$1.80 for input plus US$1 for output. The same token volumes imply US$5.60 afterwards. This is our calculation, excluding other charges; it is not a measured typical job.

For a team running that identical workload 100 times, the corresponding token subtotal would be US$280, then US$560. The point is not that either price is necessarily expensive. It is that a pilot and its eventual operating budget can sit on opposite sides of a scheduled rate change.

The current launch describes a one-million-token unified input-output limit. It should not be read as a promise of one million input tokens plus another million output tokens. Google points developers to a continuation protocol for the relevant setup.[1]

Capability claims and the release itself

Google positions the developer release around longer, more complex assignments and the ability to work across tools. Those are vendor descriptions of capability, not an independent finding that Argon is best for every project.[2]

The immediately usable distinction is between the access channel, the assignment and the charging model. A subscriber exploring an idea in the app is not buying the same thing as a developer paying for repeated API execution. Argon’s broader release brings both into reach; 30 November is the date API users should carry into their cost assumptions.

Report an error or suggest a correction