Google limits Gemini 4 Argon to trusted cyber defenders at launch

Google on Sept. 30 unveiled Gemini 4 Argon, its next frontier model, and said the first public cohort will be a set of trusted cyber defenders in its Fairwind Program rather than a general consumer or developer launch.

Koray Kavukcuoglu, Google DeepMind SVP and Google’s chief AI architect, wrote that Argon is built for deep reasoning across long-horizon workflows in software engineering, enterprise knowledge work such as legal and finance, and cybersecurity defence. For Fairwind partners and Google’s own teams, the company said it will release Argon without cyber guardrails so defenders can use its full capability to find, validate and patch critical vulnerabilities. Google said it is also engaged in the U.S. government’s voluntary pre-release model access process while it expands access gradually.

The company said Argon is already powering internal Google workflows — including large-scale C/C++ to Rust migrations and memory optimizations across data centres — and is expanding the model’s output limit to 1 million tokens. Introductory API pricing was listed at US$2 per million input tokens and US$10 per million output tokens, rising to US$4 / US$20 after the introductory period, with broader availability expected to start with paid API customers and Google AI Ultra subscribers after further safeguard work on misuse, prompt injection and misalignment monitoring.

The constrained rollout is Google’s bet that frontier cyber capability should reach defenders before it reaches everyone else — a posture that will be judged by how quickly Fairwind feedback hardens the model and how soon Canadian enterprises and security teams can access Argon through paid channels without waiting for a full consumer drop.

Sources: Google DeepMind blog, “Introducing Gemini 4”; The Verge.