Google announces Gemini 4 Argon, a new frontier AI model for coding and cyber defense
Google has unveiled Gemini 4 Argon, a frontier AI model built for real-world coding, enterprise knowledge work, and cybersecurity defense, though it is currently rolling out only to a limited group through the Fairwind Program.
On September 30, 2026, Google announced Gemini 4 Argon, its newest frontier model, positioned as strong at real-world software engineering, enterprise knowledge work such as legal and finance tasks, and cybersecurity defense. It follows Google's earlier Gemini models.
Who can access it now
According to Google's announcement, Gemini 4 Argon is currently rolling out only to a set of trusted cyber defenders through Google DeepMind's Fairwind Program. General consumers and businesses cannot access the model at launch.
Key highlights
- A context window of up to 1 million tokens, aimed at long, multi-step reasoning tasks.
- Promoted strengths across three areas: real-world coding, enterprise knowledge work (legal and financial), and autonomous handling of cybersecurity vulnerabilities.
When it will expand
Google says it is prioritizing rigorous safety testing before making Gemini 4 Argon broadly available, and has not given a timeline for wider release. Interested users should watch for further announcements rather than expect immediate access.