Google announces Gemini 4 Argon — limited first to trusted cyber defenders

On 30 September 2026, Google announced Gemini 4 Argon through DeepMind senior vice president Koray Kavukcuoglu, according to the company blog and coverage by Ars Technica and TechCrunch. The first rollout is limited to trusted cyber defenders in Google’s Fairwind Program, with a phased path before broader access.

What Google claimed

Google says Argon reaches frontier performance in software engineering, legal and finance knowledge work, and cyber defense, and can autonomously find, validate and patch vulnerabilities. The output token limit was raised to 1 million (from 64K). Introductory API pricing was listed at $2 per million input tokens and $10 per million output tokens, then $4 / $20. On DeepSWE v1.1, Google cited 77.9%. Wiz reportedly used Argon via Scan for Good and found a critical healthcare-software vulnerability that other frontier models missed; no public CVE details were provided in that coverage.

Access path and policy context

Broader access is expected later, starting with the paid API and Google AI Ultra. Google said it engaged in the U.S. government’s voluntary pre-release process. Public general availability was not announced on day one.

Limits of this report

Performance figures, pricing and vulnerability claims are Google’s (and partner anecdotes) as reported on 30 September 2026; independent third-party audits were not cited in that coverage. This article does not endorse the model or guarantee security outcomes.