What actually happened
On 30 September 2026 Google announced Gemini 4 Argon, the first Gemini 4 frontier model. It is aimed at long-horizon software engineering, legal and finance knowledge work, and cybersecurity defence.
The number that matters for a small shop is the introductory API price: $2 per million input tokens, $10 per million output tokens, and cached input at 95% off the input rate. After the introductory period, Google says that becomes $4 / $20. Google did not say how long the intro window lasts.
Argon’s output limit is 1 million tokens, up from 64K on previous Gemini models. That is headroom for an agent that has to keep writing through a long job instead of stopping mid-pass.
Access is not general. Argon is rolling out first to trusted cyber defenders in Google’s Fairwind Program. Next in line are paid API customers and Google AI Ultra subscribers. There is no date.
Why a small team cares
A two-person agency that already runs coding or automation agents cares about two things: whether a cheaper frontier model can finish a long job in one trajectory, and whether that price is real on Monday.
Google’s own examples are internal, not client work. Argon agents are being used on C/C++ to Rust migrations, from tens of thousands of lines up to 800K+ lines on parts of the Fuchsia Zircon kernel, with automated and manual audit before production. That is the class of job the 1M output window is built for.
On published tests, Google reports Argon first on Zapier’s AutomationBench at 51.3% (end-to-end execution across business apps) and 77.9% on DeepSWE v1.1 (long-horizon software engineering). Those are vendor-cited scores. Treat them as a reason to queue a bake-off, not as proof your client repo will move.
Hype vs useful
Hype is “frontier model, switch the stack.” Useful is the price sheet plus the access gate. You cannot call Argon on a client job today. When paid API opens, run one real job against whatever you already use (Claude Sonnet / Opus 5.5, GPT-6.1 Sol if you have it) at the then-current Argon rate. If the intro $2 / $10 is still live, it earns that test. If Google has already moved to $4 / $20, price the same job at the standard rate before you change a default.

