Google DeepMind Unveils Gemini 4 Argon with 1 Million Token Output Window

Google DeepMind has introduced Gemini 4 Argon, a new model variant featuring an unprecedented 1 million token output context window. Access is currently restricted to select government entities and cybersecurity partners participating in Google's Fairwind Program. A 1M token output window is a major breakthrough in generative AI, enabling models to generate massive codebases, extensive reports, or complex technical analyses in a single inference step. Restricting its rollout to cyber defenders highlights an industry trend toward leveraging specialized high-capacity models for critical national security and defense applications. While existing models primarily expand input context, Gemini 4 Argon focuses on output generation capacity, allowing for far longer coherent responses. The model is deployed exclusively through the Fairwind Program, which aims to give cyber defense organizations a head start against advanced AI-driven threats.