Google announced Gemini 4 Argon on September 30, 2026, as its new frontier large language model optimized for long-horizon software engineering, enterprise tasks, and cybersecurity defense, with a 1-million-token output limit and strong benchmark results exceeding rivals like Claude Opus 5.5. The model entered a phased rollout immediately, starting with vetted cybersecurity partners via the Fairwind Program and internal Google teams while undergoing U.S. government pre-release safety reviews. Traders are monitoring the timeline for expansion to paid API users and Google AI Ultra subscribers, as Google has signaled broader availability will follow quickly once guardrails are validated. This controlled approach, driven by capability and safety considerations, shapes near-term market-implied odds around initial public or developer access.