Google announced Gemini 4 Argon on 30 September 2026, with initial access limited to selected cybersecurity defenders. For developers considering an evaluation, the immediate distinction is between the model Google has described and the access it has actually opened.
Early access runs through Fairwind
In its Gemini 4 Argon announcement, Google says it is rolling the model out to trusted defenders through its Fairwind Program. The company plans wider availability for developers, enterprises and consumers, starting with paid API customers and Google AI Ultra subscribers, but gives no release date.
The Fairwind Program page says a set of partners gets exclusive Argon access and can use it with CodeMender for vulnerability research and patching. That is a controlled early deployment, not a general developer release. Google's descriptions of coding and defensive security performance are company claims, not results from a GeethanTech test.
Output limit and prices are announced
Google says Argon will support up to one million output tokens, compared with a previous 64,000-token output limit. This is a limit on generated output, not evidence that every long response will be useful or reliable. The announcement does not establish an input-context limit for Argon.
Google lists introductory API prices of $2 per million input tokens and $10 per million output tokens. After the introductory period, it says those rates become $4 and $20. Google has not specified when that period ends, and the price announcement does not itself mean regular developers can call the model today.
What developers can assess now
Google cites internal coding projects and model evaluations across software engineering, knowledge work and security. Those examples indicate the workloads it is targeting, but they do not replace independent testing on a team's own tasks. A practical pilot must wait for confirmed access, published API terms and a measured fit for the workload.
