Google released Gemini 4 Argon on 30 September, describing it on the DeepMind blog as "our next era of frontier intelligence" — the company's most capable model to date and, according to the Vals AI benchmarking index, the current leader across frontier model evaluations, outperforming OpenAI GPT-6 Astra and Anthropic's Fable and Opus.
Argon is engineered specifically for complex long-horizon workflows. Its headline capability is autonomous security operations: the model can independently identify, validate, and patch critical software vulnerabilities across large codebases without human sign-off at each step. It also handles long-video visual analysis, chart interpretation, and multi-step reasoning tasks of the kind that typically require human review cycles. Google's own engineering teams are already using Argon internally across daily tasks.
The launch strategy departs from the standard broad-release pattern. Argon is available initially only to selected partners under Google's Fairwind Program — a proactive cyber defence initiative serving governments and enterprises — with no public pricing or general availability date disclosed. The Fairwind gating mirrors the controlled rollout OpenAI used for GPT-6 Astra's initial operator access and signals a deliberate enterprise-first deployment approach for Google's highest-capability tier. For enterprise buyers in financial services, healthcare, and critical infrastructure, the autonomous vulnerability-patching capability is the most significant enterprise AI announcement of this cycle.
Separately, Google confirmed that the Gemini app has reached one billion monthly active users — matching ChatGPT's stated figure and establishing, for the first time, a genuine two-player market at consumer AI scale.
Gemini 4 Argon generated the largest Google Trends spike in AI model search traffic for the week of 29 September – 1 October 2026.