Google announced Gemini 4 Argon on September 30, 2026, positioning it as a frontier large language model optimized for long-horizon software engineering, enterprise tasks like legal and finance work, and cybersecurity defense. The model launched with a 1-million-token output limit and posted leading benchmarks, including 77.9% on DeepSWE v1.1 ahead of Claude Opus 5.5 and GPT-6 Astra. Access began immediately but remains gated: initial rollout targets trusted partners in Google's Fairwind Program for cyber defenders, followed by phased expansion to Google AI Ultra subscribers and paid API users after U.S. government safety reviews. No public API or general availability date has been set, leaving trader focus on the speed of Google's phased rollout amid competitive pressure from OpenAI and Anthropic.