Gemini 4 Argon reached general availability, at $2 until December
Google's answer to a fortnight of price cuts went generally available on 13 October. It undercuts the competition until 1 December, and then it doubles.
Gemini 4 Argon reached general availability in the Gemini API, two weeks after Google announced it on 30 September and sent it first to cyber defenders through the Fairwind Program. The introductory price is $2 per million input tokens and $10 per million output, with cached input discounted 95%. On 1 December that becomes $4 and $20. The context window is 1M tokens. The benchmark picture is split. The Vals Index ranks it first of 41 models at 68.90%, ahead of Claude Sonnet 5.5, Opus 5.5 and Fable 5.1, and Google's own table gives it 77.9 on DeepSWE v1.1 against 74.2 for Opus 5.5. But it scores 55.0 on FrontierSWE v2 where GPT-6 Astra scores 65.5, and Artificial Analysis puts its intelligence index at 52.6 against 57.6 for Opus 5.5.
Why this one is different
Everything else released this autumn has competed by being cheaper than the model above it. Argon competes by being cheaper than itself, later. A price that doubles on a published date is a different instrument from a discount: it buys adoption now and books the revenue in December, and it puts the switching decision in the customer's calendar rather than the vendor's. Google has done this twice before, with Gemini 3.7 Flash and 3.8 Flash, and both times the introductory rate was the headline and the standard rate was the footnote.
The price is real until 1 December. Then it is double.
How we got here
- 13 Aug 2026Gemini 3.7 Flash ships at an introductory price that doubles on 1 January 2027.
- 2 Sep 2026Gemini 3.8 Flash arrives at $0.75 per million input tokens.
- 30 Sep 2026Google announces Gemini 4 Argon, to cyber defenders first.
- 13 Oct 2026Argon reaches general availability at $2 and $10, until 1 December.
What it does and does not mean
The two independent numbers disagree with the vendor's. Vals ranks Argon first overall while Artificial Analysis puts it five points behind Opus 5.5, and both are measuring something real; which one matches your work is an empirical question about your own tasks, not a thing a leaderboard can settle. The pricing deserves the same care. At $2 it is competitive with everything released in the past month. At $4 it is the most expensive of them, and that change is six weeks away and already published.