Google announced Gemini 4 Argon on September 30, 2026, as its new frontier large language model optimized for long-horizon software engineering, enterprise knowledge work, and cybersecurity defense. The model achieved leading scores on benchmarks including DeepSWE v1.1 at 77.9 percent and the Vals Index, with an expanded 1 million output token limit. Initial rollout targets vetted participants in the Fairwind Program for cyber defense testing, followed by paid API users and Google AI Ultra subscribers. Broader developer and consumer access remains unscheduled pending phased safety evaluations and guardrail iterations. This limited launch after skipping the earlier Gemini 3.5 Pro has shaped trader views on near-term availability timelines.