Google unveils frontier Gemini 4 Argon AI model, begins rollout to cybersecurity partners

Google on Wednesday unveiled Gemini 4 Argon, a new frontier artificial intelligence (AI) model, beginning its rollout to trusted cybersecurity defenders ahead of wider availability.

The model is designed to sustain deep reasoning across complex tasks in software engineering, legal and financial work, and cybersecurity, it said.

“Today, we’re announcing our new frontier model, Gemini 4 Argon,” Koray Kavukcuoglu, Google DeepMind’s senior vice president and Google’s chief AI architect, wrote.

Initial access is being provided through Google’s Fairwind Program. The company is also participating in the US government’s voluntary process for pre-release model access.

Google plans to make Argon available to developers, enterprises and consumers “as soon as possible” while gathering feedback and refining safeguards.

The company said Argon can autonomously identify, validate and patch critical software vulnerabilities.

Google is expanding the model’s output limit to 1 million tokens from 64,000, allowing it to sustain longer reasoning and generation.

Thousands of Google employees are already using Argon for specialized coding, research and writing, the company said.