Gemini 4 Argon explained: what it means for you

Google has made Gemini 4 Argon, its top model, available to paying users and developers worldwide. Until now, only selected test partners could use it.

Google opened Argon on 14 October, just under two weeks after introducing it, according to heise online. Other outlets date the rollout to 13 October. The introduction date of about two weeks earlier was not independently checked.

Here is who gets it, and where:

Up to 10,000 selected scientists can use Antigravity's highest usage tier free for six months. They must apply to get it.

Argon has a context window of one million tokens. A token is a small chunk of text, and the context window is how much the model can handle at once. Here, that total covers your input, the model's internal calculations and its output.

Artificial Analysis (AA), an independent testing group, has benchmarked the model. On its Intelligence Index, Argon scores 53 at its highest reasoning tier. Claude Opus 5.5 (Max) leads with 58. Claude Sonnet 5.5 (Max) and Meta's Muse Wonder Full (Max) each score 56.

Argon does better on other tests. On AutomationBench-AA, it met 78 percent of task goals without breaking rules. Claude Opus 5.5 (Max) managed 70 percent and Gemini 3.8 Flash (High) managed 60 percent.

On AA-Omniscience, a test of factual accuracy, Argon answered 50 percent correctly. It was clearly wrong on only 15 percent of the remainder, often giving no answer or a partial one instead. AA calls its hallucination rate, meaning how often it states false things, comparatively low.

If you pay for Google AI Pro or Ultra, you can now use Argon in the Gemini app and in Search's AI mode. If you use the free tier, the notes don't say whether Argon is available to you.

If you build with AI, the introductory API prices are $2 per million input tokens and $10 per million output tokens. Cached input tokens, which are text the model has already processed, cost 95 percent less. The prices double from 1 December.

AA put the average cost of one task on its Intelligence Index at $1.99 at the introductory prices, or about $3.98 without the discount. AA said: "The comparatively low costs are primarily due to the temporarily favorable token prices".

In short, Argon is not the top scorer on AA's general index. It does well on following rules in tasks and on avoiding wrong answers.

The introductory prices end on 1 December, when they double. Scientists who want the free Antigravity tier need to apply. The notes don't say how or by when.