Gemini 4 Argon is out: who gets it and what it costs

Google released Gemini 4 Argon on October 13, its first major frontier model since Gemini 3 in late 2025. It replaces the Gemini Pro line at the top of Google's range, and Google says it beats OpenAI's GPT-6 Astra and Anthropic's Claude Fable 5.1 and Opus 5.5 on a range of benchmarks in coding, finance, law, cybersecurity and science. Whether you can use it today depends on which Google plan you pay for.

Who can use Argon today

• Google AI Pro and Ultra subscribers get it in the Gemini app and in AI Mode in Search. Google's plan page now lists Argon under both plans.
• Google AI Ultra subscribers also get it in Antigravity, Google's agentic coding platform.
• Developers and businesses get it through the Gemini API in AI Studio and the Gemini platform.
• Free users and Google AI Plus subscribers do not. Plus lists Gemini 3.6 Flash instead, and Google has not said when, or whether, Argon comes to those tiers. Google says more Gemini 4 models are coming.

In the US, Google AI Pro is $19.99 a month and Ultra starts at $99.99 a month, so Pro is the cheapest way to try Argon in the Gemini app. Our Gemini review covers what else each plan includes.

What it costs through the API

Argon launches at an introductory price of $2 per million input tokens and $10 per million output tokens, with cached input 95 percent cheaper. After November 30, 2026, the price rises to $4 per million input tokens and $20 per million output tokens. The Verge puts the launch price at around a fifth of what OpenAI and Anthropic charge for GPT-6 Astra and Fable 5. At the full price, Argon's input cost matches Claude Opus 5.5's $4 and stays well under the $10 per million that GPT-6 Astra and Fable 5.1 charge.

If you plan to build on Argon, budget for the November 30 price, not the launch one. Google also lists a 1 million token limit covering input and output together through the API.

How far to trust the benchmarks

The headline claims are Google's own. Argon ranks first on the Vals Index, which scores models on finance, coding, legal and tax work, and Google reports a 0.7 percent success rate for attackers on the Gray Swan indirect prompt injection test. In an earlier round of that test, GPT-6 Astra scored 8.5 percent and Claude Opus 5 4.8 percent. The Decoder points out that Google's write-up does not make clear whether the same tests were run for each model, and that Meta's Muse Wonder Full, released the day before Argon, now sits above it in some overall rankings.

Google's internal numbers are bolder still: it says Argon raised agent-written code submissions by more than 35 percent and cut rollbacks by nearly a third. Those are company figures, so treat them as a claim to test on your own work.

Where else you will find it

Third-party tools are adding Argon quickly. Cursor's model list already includes Gemini 4 Argon at the same $2 and $10 per million token rates, next to Claude Sonnet 5.5 at the same price. That gives coding teams a quick way to compare the two on real tasks before the price changes.

Google is also offering 10,000 scientists six months of Antigravity's highest usage tier free of charge, chosen on a rolling basis. Applications are open on Google's site.

Should you switch?

If you already pay for Google AI Pro, Argon costs you nothing extra, so try it on the hardest work you do and compare it with what you use now. If you are on Plus or the free plan, nothing changes yet. Developers get a cheap window until November 30 to test it; after that it is still cheaper per token than GPT-6 Astra or Fable 5.1, but no longer a fraction of their price.

Sources: Google, The Verge, The Decoder, Google AI plans, Cursor models and pricing.