Google unveils Gemini 4 Argon, its new frontier model, with limited access for now
IA en un minuto newsroom · Editor: Jon Elgezabal

In 30 seconds
Gemini 4 Argon, the new frontier model from Google, can write up to 1 million tokens in one answer, compared with a previous cap of 64,000. For now only selected cybersecurity defenders in the Fairwind program can use it, and Google has not said when others will get access. The introductory price is $2 for each million input tokens and $10 for each million output tokens.
Google presented Gemini 4 Argon, its new frontier model, on September 30, though access is limited for now.
Argon is built for long and complex tasks: programming, corporate work in fields such as legal and finance, and cybersecurity defense. Its output limit, meaning how much the model can write in one response, grows to 1 million tokens from the previous 64,000.
At this stage it reaches only a group of trusted cybersecurity defenders, by way of Google's Fairwind program. The company is releasing it in phases and says developers, businesses and consumers will get it as soon as possible, with no date given. The introductory price will be $2 per million tokens of input and $10 per million of output.
After that introductory period the price will go up to $4 per million input tokens and $20 per million output tokens, while cached input tokens will cost 95% less. Google adds that when it opens Argon more widely it will begin with paying customers of its API and subscribers to Google AI Ultra. In Google's own tests, Argon reaches 77.9% on DeepSWE v1.1, a benchmark of long, real-world programming tasks.
Why it matters · analysis and opinion
Who is Gemini 4 Argon for today? Only for a group of cybersecurity defenders chosen by Google; everyone else will have to wait, with no date. What sets it apart from what came before is how much it can write in one go, a leap designed to deliver long programs or reports in a single response instead of splitting them up. Starting with cyber defense suggests that Google sees sensitive capabilities in it and prefers to measure before opening up. For teams planning to use it for coding or for legal and financial work, one detail counts in the budget: the launch price is temporary and then doubles, although caching makes repeated input much cheaper. And the performance result that comes with the announcement is from Google's own tests.
Official source: Google · Written with the help of AI: how we make the news


