Google

Artificial Intelligence · Cybersecurity · Enterprise Software

Launch

Google launches Gemini 4 Argon, its most advanced frontier AI model

The phased rollout to vetted cyber defenders reflects rising safety fears after rival AI agents have started misbehaving in the wild.

• Google unveiled Gemini 4 Argon on September 30, 2026, its new frontier AI model, rolling it out first to a small group of trusted cybersecurity partners through its Fairwind Program.
• Argon is built for deep reasoning across long, complex tasks in software engineering, enterprise knowledge work like legal and finance, and cybersecurity defense.
• The model's output limit jumps to 1 million tokens, up from 64K, letting it reason through hundreds of thousands of tokens in a single trajectory for tougher problems.
• Argon will launch at an introductory price of $2 per million input tokens and $10 per million output tokens, with cached inputs discounted 95%.
• Across 18 disclosed benchmarks, Argon leads outright on 12 and ties for first on one, outscoring GPT-6 Astra and Claude Opus 5.5 on coding tests like DeepSWE v1.1 (77.9% vs. 74.1% and 74.2%).
• Gemini 4 arrives nearly a year after Gemini 3, following the cancellation of a planned Gemini 3.5 Pro release and a year spent focused on cheaper Flash models.
• Google is engaging the U.S. government's voluntary pre-release safety process and plans to expand access to paid API customers and AI Ultra subscribers before a full public release.
• The cautious, staged release underscores how frontier AI labs are now weighing capability gains against real incidents of AI agents misbehaving, reshaping how models reach the public.