Google Debuts Gemini 4 Argon With 1M Token Output, Rolling Out to Cyber Defenders First

Google released Gemini 4 Argon on September 30, 2026, a frontier AI model designed to tackle complex reasoning tasks in software engineering, enterprise knowledge work, and cybersecurity defense. The model supports up to 1 million output tokens—a significant jump from the 64,000-token limit of earlier Gemini versions—enabling the system to reason through complex problems without interruption.