Google Unveils Gemini 4 Argon AI Model for Cybersecurity and Development

Google announced its new AI model, Gemini 4 Argon, on September 30, 2026. The model is initially available to trusted cyber defenders in Google's Fairwind Program. Google plans to expand access to developers, businesses, and consumers, but has not given a public release date.

Argon claims to improve upon previous models with its ability to help with coding, research, business tasks, and cybersecurity. It can supposedly produce up to 1 million output tokens, compared to the 64,000-token limit of previous models. However, Google has not explained how this limit would work in a consumer app.

Google reports that Argon leads its rivals on certain tests, including the Vals Index Knowledge Work and AutomationBench Knowledge Work benchmarks. The company has announced API pricing for developers and businesses, with introductory rates of $2 per 1 million tokens for input and $10 per 1 million tokens for output.

While Google's scores are promising, independent testing is still needed to confirm the model's capabilities. For now, it seems that Argon will not be available to consumers until later this year or in early next year at the earliest.