On September 30, 2026, Google announced Gemini 4 Argon, a new frontier model. Google says it showed strong performance in real-world software development, enterprise knowledge work such as law and finance, and cyber defense.
Rollout order
Gemini 4 Argon is not being made available to the public right away. According to the official blog, it will roll out in the following order.
- First, to cyber defense experts and organizations trusted by Google, through Google's Fairwind Program
- Next, starting with paid API users and Google AI Ultra subscribers, expanding to developers, businesses and consumers "as soon as possible"
Given that the model is capable enough to find and fix vulnerabilities, Google appears to have chosen to put it in defenders' hands first.
Pricing
During the introductory period, API pricing is $2 per million input tokens and $10 per million output tokens, with a 95% discount on cached input. Once the introductory period ends, pricing is set to move to the standard rate of $4 for input and $20 for output.
Key benchmarks
These are the main figures Google published.
- 77.9% on DeepSWE v1.1, an evaluation of real-world software development
- 51.3% on AutomationBench, a business automation evaluation (ranked first)
- 91.7% on LVBench, which measures long-video understanding
- 68% on CWE-bench v1, which measures vulnerability fixing (tied for first)
Google says the model also ranked first on the Vals Index, which spans finance, coding, law and tax, and on Gray Swan's benchmark of resistance to indirect prompt injection (IPI). All figures come from Google's own announcement.
CoAI's take
For now, this is not a model that everyday Gemini app users can try. Google AI Ultra subscribers, however, are likely to get access relatively early, suggesting a pattern of offering the most capable models on higher-tier plans first.
