Gemini 4 Argon Arrives, but Only Cyber Defenders Get It First

Google DeepMind announced Gemini 4 Argon on 30 September 2026, describing it as a frontier model for real-world software engineering, enterprise knowledge work such as legal and finance, and cyber defence. Unusually, the first people to get it are not developers or consumers. Argon is rolling out to a set of trusted cyber defenders through Google's Fairwind Program.

Google says a phased release is the point. It is taking part in the US government's voluntary process for pre-release model access while it gradually widens availability, and it will open Argon to developers, enterprises and consumers "as soon as possible". There is no public date yet.

The pricing is public, though. Argon launches at an introductory $2 per million input tokens and $10 per million output tokens, with cached input tokens priced at 95 percent off the input price. The model carries a 1 million token limit, which matters for the long-running agent work Google keeps pointing to.

The most interesting section is what Google says Argon is already doing internally. Its quantum computing researchers used it to optimise the qubits-times-gates cost of a subroutine, and in one example it beat the published baseline by 40 percent in a matter of minutes. A team of Argon agents analysed fleet-wide profiling telemetry and applied memory optimisations across Google's data centres, freeing over 300 TiB once rolled out, with an estimated 500 TiB to 1 PiB in total.

Argon agents are also migrating C and C++ codebases to Rust, from core libraries such as re2 and libgav1 up to more than 800,000 lines for the Fuchsia Zircon kernel. Google is careful to say those rewrites are going through automated and manual auditing and emulation testing before they reach production.

For a studio like ours, the practical read is simple. Treat the benchmark claims as Google's own until independent testing lands, and plan around the cached-token discount if you build anything that re-reads the same large context. We will test it properly once API access opens.

The one thing to watch is the access model. A frontier release that starts with defenders and government pre-release review is a new pattern, and it may become the norm for the most capable models.