Google announced Gemini 4 Argon on September 30, 2026, as its new frontier large language model optimized for long-horizon tasks in software engineering, enterprise knowledge work, and cybersecurity defense. The model features a 1-million-token output limit and leads or ties rivals including Claude Opus 5.5 and GPT-6 Astra on most Google-reported benchmarks such as DeepSWE v1.1 and Vals Index. Rollout began immediately but is limited to internal teams and vetted participants in the Fairwind Program for cyber defenders, with Google engaging U.S. government pre-release reviews before expanding. Broader paid API access and consumer availability via Google AI Ultra are planned next without a firm date, following the skipped Gemini 3.5 Pro. This phased safety-focused approach and strong benchmark positioning shape trader views on release timing and attribution.