Google's Gemini 4 Argon is priced at $2 and $10 per million tokens, half of Claude Opus 5.5's list price

Alphabet unveiled Gemini 4 Argon on Wednesday and called it its most capable AI model so far, with gains in coding, cybersecurity and professional work such as finance and law. Google is starting with a limited release. In its announcement, the company said Argon is going first to "trusted cyber defenders" in its Fairwind Program, while it takes part in the U.S. government's voluntary pre-release testing process. Developers, businesses and consumers will get access "as soon as possible," which is not a date.

Alphabet's Class A shares closed up 0.93% at $344.08, according to Nasdaq data. In after-hours trading they were at $348.62 at 6:51 p.m. ET, up another 1.32%. CNBC reported that the launch came a day after CEO Sundar Pichai signed a voluntary AI safety agreement with President Donald Trump and other tech executives at the White House.

The number behind the number: the price

Most coverage focused on benchmarks. The detail that matters more for anyone paying an AI bill is the price, which Google gave in a footnote. Argon will launch at an "introductory price" of $2 per million input tokens and $10 per million output tokens. Cached input, meaning text the model has already seen and is reusing, costs 95% less, or $0.10 per million.

That is exactly half of Anthropic's list price. As we reported on September 22, Claude Opus 5.5 costs $4 per million input tokens, $20 per million output tokens and $0.20 per million cached reads.

Here is how that works out on one agent job that reads 5 million tokens, 4 million of them cached, and writes 500,000:

Two caveats. "Introductory" means the price can change, and Google did not say for how long it will last. And nobody outside the early-access group can buy it at any price yet.

What the headline gets wrong: a lower rate can still mean a bigger bill

Google raised Argon's output limit to 1 million tokens from 64,000, so the model can work through a problem for much longer in a single response. That is the feature it is selling. It also changes what a single response can cost. At $10 per million, one response that uses the full limit costs $10 in output tokens. The old 64,000-token cap would have held the same response to about $0.64. So a lower price per token does not guarantee a lower bill. Anyone who puts Argon to work on long tasks should set their own caps on how much it can write.

What Google says it can do

• Coding: 77.9% on DeepSWE v1.1, which Google calls a new state of the art for long software engineering tasks. Google says Argon agents are rewriting C and C++ code in Rust across the company, including more than 800,000 lines of the Fuchsia Zircon kernel.
• Business work: first place on Zapier's AutomationBench, at 51.3%, and first on the Vals Index, which weights finance, coding, legal and tax work by each sector's share of U.S. GDP. CNBC reported that Argon leads OpenAI's GPT-6 Astra and recent Anthropic models on that index.
• Cybersecurity: tied for first at 68% on a vulnerability-fixing benchmark. CNBC said it tied GPT-6 Astra and Grok 4.7 on cybersecurity tests.
• Google's own data centers: Argon agents found memory savings that freed more than 300 TiB once rolled out, with 500 TiB to 1 PiB expected in total. Google did not put a dollar value on that.

All of these are Google's own figures. Independent test results will not exist until more people can use the model.

Who it actually hits

Small software shops and agencies that resell AI features on per-token contracts will face customers asking why their price has not halved. The answer for now is that Argon is not generally available. Businesses deciding on a model today can take this as a sign of where prices are heading, not something they can buy yet. Alphabet shareholders should keep an eye on the gap between launch and general availability, because revenue starts when enterprise customers can buy, not at the announcement. Our Alphabet explainer covers what else is driving the stock.

Sources: Google; CNBC; Nasdaq market data; Anthropic pricing as reported by The Company Chronicle. The cost table is an illustration using list prices, not a quote. This is market information, not investment advice.