Google DeepMind released Gemini 4 Argon on September 30, a frontier model that raises the output limit from 64,000 to 1 million tokens. But almost nobody can use it today: the first group with access is trusted cyber defenders.
According to Google's official announcement, Gemini 4 Argon is built to sustain deep reasoning across long, multi-step workflows: real-world software engineering, enterprise knowledge work (legal and finance), and cyber defense. Its most striking feature is a 1-million-token output limit, up from 64,000 in the previous generation, which allows hundreds of thousands of tokens to be generated in a single pass on long problems.
Argon is rolling out first through the Fairwind Program, which brings together trusted cyber defenders (governments and critical-infrastructure operators, according to press coverage). For that group, Google offers the model without its cybersecurity safeguards so it can be used for autonomous vulnerability discovery and patching. Paying API customers and Google AI Ultra subscribers come next, followed by the general public, with no confirmed date. Google says the launch ships with defenses against misuse and prompt injection, misalignment monitoring, and isolated testing environments.
These are benchmarks reported by Google in its own announcement; it's worth waiting for independent evaluations before treating them as definitive.
Carlos Montiel is an enterprise AI solutions architect. He implements LLMs, Agents, RAG and orchestrators for companies across Guatemala and Latin America. Reach out for a consultation.
Contact Carlos Montiel