Google Introduces Gemini 4 Argon With 1M Token Capacity and Frontier Performance in Engineering Workflows

Google DeepMind has launched Gemini 4 Argon, a frontier AI model featuring an industry-leading 1 million token output limit and strong performance across software engineering, enterprise knowledge work, and cybersecurity defense. The model was announced September 30, 2026, and represents a 16-fold expansion of output capacity from the previous 64,000 token limit, enabling extended reasoning trajectories and complex multi-step workflows within a single execution.