Google made Gemini 4 Argon, its newest flagship reasoning model, broadly available on October 13, 2026. Paying developers can now call it through the Gemini API as a stable model, and Google AI Pro and Ultra subscribers can use it in the Gemini app. Google first announced Argon on September 30, when access was limited to a small group of testers and cyber defenders.
Key details
- Model ID:
gemini-4-argon, listed as “Stable” in the Gemini API model list. - Context: a unified 1M-token window, with long generations of up to 1M output tokens.
- Consumer access: Gemini app for Google AI Pro and Ultra subscribers. Free accounts and Google AI Plus do not get it.
- Developer access: paid API tier only. Google’s pricing page states that Argon “is not available on the Free tier or on the Google AI Pro plan”, so a Pro subscription does not cover API use.
- Caching: implicit caching only; explicit caching is not supported.
What it costs
The launch prices are temporary. Standard API rates double on December 1, 2026:
| Per 1M tokens (Standard) | Until Nov 30, 2026 | From Dec 1, 2026 |
|---|---|---|
| Input | $2.00 | $4.00 |
| Output (incl. thinking) | $10.00 | $20.00 |
| Cached input | $0.10 | $0.20 |
Batch and Flex requests cost half the Standard rate ($1.00 input and $5.00 output until November 30, then $2.00 and $10.00). The Priority tier costs more: $3.60 input and $18.00 output now, rising to $7.20 and $36.00.
How it compares
Even after the December increase, Argon’s list price matches Claude Opus 5.5 ($4 input / $20 output per 1M tokens) and stays well below GPT-6 Astra ($10 / $50 for prompts up to 272K tokens). Until November 30 it is half the price of Opus 5.5.
On quality, the picture depends on who is measuring. Google’s own launch table puts Argon ahead on most rows, including 77.9% on the DeepSWE v1.1 coding test and 68.9% on the Vals Index for knowledge work. It does not lead everywhere: Google’s table shows it behind Claude Opus 5.5 on Terminal-bench 4.0. These are vendor-reported, best-case numbers run at high thinking settings, so treat them as a starting point rather than a verdict.
What it means for you
- Developers: the next six weeks are the cheapest time to test Argon on your own coding and document workloads. Budget for the December 1 price change before you move production traffic.
- Teams with long inputs: the 1M-token window and long output limit suit large codebases, contracts and research documents.
- Consumers: you need Google AI Pro or Ultra to use Argon in the Gemini app. Free and Plus users stay on lighter models.
- Anyone without a paid plan: there is no free API access, so prototyping requires billing to be enabled.
For full specs, see our Gemini 4 Argon model page, and compare options in best AI for coding and best chat assistants.
