Google Announces Gemini 4 Argon: 1M-Token Output, First Released to Fairwind Cyber Defenders

Koray Kavukcuoglu, SVP of Google DeepMind, announced Gemini 4 Argon on September 30, 2026 as the next generation of Gemini frontier models. Google says it sustains deep reasoning across complex, long-horizon workflows spanning real-world software engineering, enterprise knowledge work such as legal and finance, and cybersecurity defense. The output limit rises to 1M tokens, up from 64K in the previous generation.

Google's published results include 77.9% on DeepSWE v1.1, 68% on the CWE-bench v1 vulnerability-remediation benchmark, 51.3% on AutomationBench for business automation, and 91.7% on LVBench for long-video understanding. These are vendor-reported numbers that still need independent validation on real workloads.

The release model is the key detail. Argon is currently rolling out only to trusted cyber defenders through the Fairwind Program. Google says it will gather feedback from early testers and iterate on guardrails before making Argon available to developers, enterprises, and consumers as soon as possible, starting with paid API customers and Google AI Ultra subscribers. Google also says it is participating in the U.S. government's voluntary pre-release model access process.

Announced introductory pricing is $2 per million input tokens and $10 per million output tokens, with cached input 95% off; after the introductory period it moves to $4 input and $20 output per million tokens. The announcement does not say when the introductory period ends or give a date for general API availability.

For enterprises running a private AI platform or model gateway, three preparations make sense now. First, model both Argon's introductory price ($2 input / $10 output per million tokens) and its post-introductory price ($4 / $20), and set per-request output caps and departmental quotas with the 1M-token output limit in mind. Second, estimate what the 95% cached-input discount actually saves on long-document and fixed-system-prompt workloads. Third, prepare an internal eval set covering code changes, legal or financial documents, and security patching, so that when general paid API access opens the model can be compared against existing cloud models rather than judged on vendor benchmarks alone.

Three things to watch: when general paid API access opens, whether Argon lands in Gemini Enterprise and Gemini Enterprise Agent Platform at the same time, and how the 1M-token output behaves in real cost and latency. Until general availability, Argon is a roadmap signal for enterprises rather than a model they can procure today.