Google announced Gemini 4 Argon on September 30, 2026, positioning it as a frontier large language model optimized for long-horizon software engineering, enterprise knowledge work, and cybersecurity defense, with an industry-leading 1-million-token output limit and strong benchmark results such as 77.9% on DeepSWE v1.1. The model entered a controlled rollout exclusively to vetted cyber defenders via the Fairwind Program and internal teams, with Google citing safety evaluations, guardrail iteration, and U.S. government pre-release processes as reasons for the phased approach before expanding to paid API customers, Google AI Ultra subscribers, and broader users. This measured timeline, combined with the emphasis on responsible deployment amid ongoing AI safety discussions, underpins trader consensus around near-term dates while leaving room for delays if feedback reveals issues or regulatory steps intervene. No confirmed broader availability date has emerged in the first week post-announcement.