Google announced Gemini 4 Argon on September 30, 2026, as its new frontier large language model optimized for long-horizon software engineering, enterprise workflows, and defensive cybersecurity, with a 1-million-token output limit and strong benchmark leads on tasks like DeepSWE v1.1. The company immediately began a controlled rollout to vetted users in its Fairwind Program and internal teams while iterating on guardrails and engaging U.S. government pre-release processes, with no public timeline beyond “as soon as possible” for paid API customers and Google AI Ultra subscribers. As of mid-October, access remains gated and the model is absent from the general Gemini API and consumer apps. Market-implied odds heavily favor an October 13 resolution because traders interpret the two-week phased window and recent internal references as signals that broader availability could arrive imminently, though any delay in safety or regulatory steps could shift probabilities toward later dates.