Gemini 4 Argon matches its rivals on price, but only while the introductory rate lasts

Gemini 4 Argon pricing starts at $2 and $10 per million tokens, then doubles. Priya Chen shows what a 1 million token output costs and what to test first.
A single straight road running toward distant mountains in a desert.

We lined up three rate cards and got a tie: $2 per million input tokens and $10 per million output for Google’s Gemini 4 Argon, OpenAI’s GPT-6.1 Sol and Anthropic’s Sonnet 5.5. Then we read the footnotes. Only one vendor says its price will double. Gemini 4 Argon pricing is a tie for about a month, so budget at the later rate and cap output length before anyone on your team switches it on.

Key points

  • Google’s pricing page dates the change: $2 and $10 through November 30, 2026, then $4 and $20 from December 1. Cached input is 95% cheaper.
  • One maximum-length run (500,000 tokens in, 1,000,000 out) costs about $11 now and $22 later, by our arithmetic.
  • Output volume moves the bill more than the rate card, and your staff control it.
  • The benchmark wins are Google’s own, on tests Google chose.

The Gemini 4 Argon pricing tie, side by side

Google’s launch post puts Argon at the same sticker as its two rivals. Here are the three cards, with Argon listed twice:

Model Input Output
Gemini 4 Argon, to Nov 30, 2026 US$2 (about C$2.79) US$10 (about C$13.96)
Gemini 4 Argon, from Dec 1, 2026 US$4 (about C$5.58) US$20 (about C$27.92)
OpenAI GPT-6.1 Sol US$2 (about C$2.79) US$10 (about C$13.96)
Anthropic Claude Sonnet 5.5 US$2 (about C$2.79) US$10 (about C$13.96)
Prices per million tokens, standard API rate, as of Oct 2026. C$ at the Bank of Canada rate of 1.3960 on October 30, 2026. Sources: Google OpenAI Anthropic

Sol’s price applies to inputs up to 272K tokens. Anthropic lists no end date for Sonnet 5.5’s rate. Argon’s introductory rate is the only one with an expiry, and Google has already published it.

Advertisement

Access is the other catch. Google says Argon goes first to trusted cyber defenders through its Fairwind Program, with paid API customers and Google AI Ultra subscribers next, after more testing. Most small firms can’t buy it yet. We made a similar point about Claude Opus 5.5’s price cut, where the effort setting mattered more than the list rate.

We costed one long run and one ordinary month

Our own arithmetic; we measured nothing. A job that reads 500,000 tokens and writes 1,000,000 costs $1.00 for input and $10.00 for output now, so about $11. After December 1 it’s $2.00 and $20.00, about $22. Output is where the bill sits.

Earlier Gemini generations stopped at 64,000 tokens per response. Argon can write 1 million in one run, which suits rewriting a codebase. Cap the same job at the old ceiling and output costs about $0.64 now, $1.28 later. Ten full-length runs a week come to $110 now and $220 later.

Now a small team’s month, with round numbers we picked: 5 million tokens in, 2 million out. That’s $30 at the introductory rate and $60 after. Small money. Then let long outputs become routine. If the same team writes 20 million output tokens a month, output alone is $200 now and $400 later. The doubling is fixed. Volume is the variable, and it’s the one your staff control.

So the cap matters more than the rate card. Nobody on your team negotiates Google’s price, and everybody on it can accept the default length. Set the limit in the settings before handing out keys. Per-token prices can mislead the other way too, as GPT-5.5 pricing shows when you price by the finished task.

Can you trust Google’s benchmark table?

Google reports 77.9% on DeepSWE v1.1, 91.7% on the LVBench long-video test, 51.3% on Zapier’s AutomationBench for the top spot, and 68% on CWE-bench v1 for vulnerability fixes, tied for first. Google picked the tests and ran them. It led with security, automation and long-running agent work, where public evaluations are thinnest. Nobody outside Google has checked the results yet, and we wouldn’t buy on them.

A Bloomberg report relayed by Android Headlines adds a caution: some Google employees with model access reportedly find Argon strong on benchmarks and weaker on messy real-world coding and front-end design. Google denies any gap. Read the table as a list of what Google is proudest of. Your tasks are probably duller than a vulnerability benchmark, and dull tasks are where models differ.

How we’d test it before committing

Take one job your team does every week. Run it through all three vendors at current prices, keep the invoices, then cost Argon again at the December rate. Measure cost per successful task instead of cost per token. Skip demo prompts.

For features rather than cost, our ChatGPT, Claude and Gemini business comparison is the better start. Pricing is half the test, and a narrow model like the Jev AI model shows the other half is what it refuses to do; we checked those claims in our look at the Jev model. Vendor-run benchmarks like Kimi K3‘s are another trap.

A fair objection: a 1 million token output is a lab spec far from a 40-person firm’s needs, and comparing prices before general availability is a distraction. Both land. A default setting or an eager employee can still get you there without anyone deciding to, and the cap is for that case. We’d be wrong if Google kept $2 and $10, or if independent tests on ordinary business code showed the benchmark wins carry over. Read every figure as correct on October 8, 2026, and likely to move.

Frequently asked questions

How much does Gemini 4 Argon cost?

Google lists $2 per million input tokens and $10 per million output tokens at launch, rising to $4 and $20 after the introductory period. Cached input is priced 95% lower.

Who can use Gemini 4 Argon today?

Google says trusted cyber defenders get it first through its Fairwind Program. Paid API customers and Google AI Ultra subscribers follow after more testing.

Archive entry dated October 1, 2026, written and fact-checked on October 8, 2026. Sources are linked on the claims they support.

For short video pricing that landed this week, see also Open Google Vids before you buy Gemini Omni API.

Total
0
Shares
Prev
Meta’s Muse is free in Canada and OpenAI’s dots reaches 4,000 apps, so set the rules first
Food vendor at a street stall checking a smartphone between customers.

Meta’s Muse is free in Canada and OpenAI’s dots reaches 4,000 apps, so set the rules first

Meta Muse is free in Canada

Next
The ChatGPT privacy findings carry no fine, so treat them as a checklist
A padlock rests on a computer keyboard.

The ChatGPT privacy findings carry no fine, so treat them as a checklist

The ChatGPT privacy findings from Canada's regulators set no fine

You May Also Like