Searches for “Gemini 4 Argon pricing” are climbing, and the headline number needs a health warning. Google’s launch pricing is $2 per million input tokens and $10 per million output tokens. A footnote in Google’s 30 September 2026 announcement says $4 and $20 apply after an introductory period it has not defined.
Two facts that matter more than the benchmark
- Access is restricted. Argon launched to trusted cyber defenders through Google’s Fairwind Program. Google says broader access for developers, enterprises and consumers will follow “as soon as possible”, starting with paid API customers and Google AI Ultra subscribers. There is no date.
- The price is a promotion. Third-party analysts, not Google, are the ones telling buyers to plan around $4/$20. Cached input is discounted by 95%, which helps repetitive workloads but not every product.
Google also cites a demonstration of migrating a C/C++ codebase of more than 800,000 lines to Rust. That is a vendor claim, and we have not seen independent verification. Pricing figures from aggregator sites should be checked against Google’s own rate sheet.
What this means in plain English
Model prices are now a moving part of any AI product’s running costs, and vendors are using introductory rates to win adoption. If your AI feature’s business case only works at $2/$10, it does not work yet. The same warning applied to the AI coding subscription changes we covered earlier this month.
So what
When a supplier quotes an AI-powered product, ask which model it assumes, whether the price is introductory, and what happens to your monthly bill if the rate doubles or the model is retired. Good architecture lets you switch models without a rebuild. That is a design decision made at the start, which is why our AI-assisted development work keeps model choice behind a single layer. If you are costing an AI product now, start a project and we will model it at the higher price.