Google introduces Gemini 4 Argon, starting with trusted cyber defenders

Why it matters

【What it is】

Google’s blog introduces Gemini 4 Argon (Sep 30, 2026; Koray Kavukcuoglu, SVP of Google DeepMind and Chief AI Architect at Google). It is a frontier model for real-world software engineering, enterprise knowledge work such as legal and finance, and cybersecurity defense.

Access starts with trusted cyber defenders in the Fairwind Program. Google describes a phased rollout and says it is taking part in the U.S. government’s voluntary pre-release access process. Wider access has no public date and starts with paid API customers and Google AI Ultra subscribers.

【Verifiable】

Introductory price: $2 per million input tokens and $10 per million output tokens, with cached input 95% off the input price. A footnote says that after the introductory period, $4 per million input tokens and $20 per million output tokens apply.

The output token limit rises from 64K to 1 million. Stated scores: DeepSWE v1.1 at 77.9% (a new state of the art), AutomationBench at 51.3% (ranked #1), LVBench at 91.7%, and CWE-bench v1 tied for first at 68%. For trusted defenders and Google’s own teams, the page says Argon will ship without cyber guardrails.

The page calls Argon the leader on the Vals Index, Vals Finance Agent v2, and Harvey’s Legal Agent Benchmark, and does not publish those scores.

【Our read】

This is a phased frontier release that starts with defenders. The $2 and $10 rates are introductory list prices. The page gives no date for wider access.