Gemini 4 Argon: Google's new model opens to cyber defenders first

Google has announced its new frontier model, Gemini 4 Argon. According to Google's announcement, the model targets long-running workflows in real-world software engineering, enterprise knowledge work such as legal and finance, and cyber defence. The Verge reports that the company has limited access at the start.

Who can use Argon right now?

For now, only selected, trusted cyber defenders. Google says it is offering the model to this group through the Fairwind Program and wants to make it available to developers, enterprises and consumers as soon as possible. No firm date was given. The company explains the phased rollout as time to strengthen safeguards against misuse, prompt injection and misalignment.

What changed technically?

The output limit rises from 64K tokens to 1M tokens, so the model can think and write at much greater length in a single pass. Google says Argon set a new high of 77.9% on the DeepSWE v1.1 software-engineering benchmark and ranks first on Zapier's AutomationBench with 51.3%. These are results Google itself shared; the source contains no independent verification.

On pricing, Google has set introductory rates of $2 per million input tokens and $10 per million output tokens, with cached input tokens 95% cheaper than the input price.

What should teams in Turkey do?

The source gives no Turkey-specific detail, so what follows is interpretation. Because the model is not yet generally available, the task today is preparation, not migration. Software teams can list the jobs they currently split up because of output limits. As the pricing is introductory, it is safer to plan costs as a range rather than a fixed figure. Security teams can follow the Fairwind Program and decide whether to apply once access terms are clear.