Google has a new frontier model, Gemini 4 Argon, and it is not going to everyone at once. The company says Argon is rolling out first to trusted cyber defenders through something it calls the Fairwind Program. A broader audience is promised, but no date is attached to that promise.
The model is built for deep reasoning across long-horizon workflows, according to the announcement. Google names software engineering, legal work, finance and cybersecurity defense as the areas it is aimed at. The release is phased, with access expanding gradually rather than opening up all at once.
Read more: Google Unveils Gemini 3.8 Flash, Adds 3.8 Flash Cyber
Google says it is taking part in a voluntary US government process that gives officials a look at frontier models before they ship. Feedback gathered from early testers will shape the guardrails before any wider release, according to the announcement. Developers, enterprises and consumers are all listed as eventual audiences, though the company gives no timing for any of them.
Pricing details are partial. Google describes the rates as introductory, listing input tokens at $2 per million tokens and output tokens at $10 per million tokens. Cached input tokens carry a 95% discount off the input token price. That is the extent of what the company disclosed about cost.
There is no general availability date. Google released no benchmark results or performance figures. It also did not disclose the model's size, parameter count or context window. No regional breakdown was provided either, so it is unclear where Argon will reach users first. No executives or spokespeople are quoted in the notice.
The launch lands in the middle of a busy stretch for Google's AI efforts. The company previewed Googlebook laptops in May, a line that replaces Chromebooks and runs the Android stack instead of ChromeOS, with Acer, Asus, Dell, HP and Lenovo as the first partners. In September, Asus announced the Googlebook 14, a 0.99kg laptop built on Gemini Intelligence that pairs with Android phones and includes 12 months of Google AI Pro, without disclosing pricing or a release date.
Google's models have also been reaching users through other companies' products. Apple rebuilt Apple Intelligence on foundation models co-developed with Google, using Gemini technology for the underlying system. Argon, by contrast, arrives first in the hands of a narrow group of security professionals, with everyone else waiting on a timeline Google has not shared.
