Google announces Gemini 4 Argon, but you can't use it yet

Google unveiled its new frontier model, Gemini 4 Argon, but unless you're a vetted cybersecurity defender, you'll have to wait your turn.

According to Google's blog post, Argon jumps to a 1-million-token output limit, up from the previous 64K limit. Google is restricting initial access to internal teams and select security partners through its Fairwind Program—and actually removing cybersecurity guardrails for those trusted defenders. Public enterprise and developer access will come later, with pricing set at $2 per million input tokens and $10 per million output tokens.

Why it matters: Argon targets heavy, multi-step engineering and research tasks that need deep reasoning across long timeframes. Google reports internal teams are using Argon agents to migrate an 800,000-line C++ kernel to Rust and rewrite video decoder components to run 2.7× faster. If those results hold up outside Google, giving AI agents enough headroom to generate massive blocks of code in one go could simplify large codebase overhauls.

Here's the gist: you can't sign up today unless you're working on defensive security programs like partner Wiz, which used Argon to catch a critical hospital software vulnerability.

For now, the rest of us will just have to watch Google's trusted defenders put the model through its paces.