Google Unveils Most Powerful Model Yet, Gemini 4 Argon

Google has unveiled Gemini 4 Argon, its most powerful model yet, designed to tackle complex workflows and long-horizon tasks. This new frontier model is a direct challenge to OpenAI and Anthropic, with aggressive benchmark claims in software engineering, knowledge work, and cybersecurity.

The model features a 1-million-token output limit, far exceeding the previous ceiling of 64,000 tokens, giving it 'room to think deeply' and generate hundreds of thousands of tokens in a single trajectory. This is particularly beneficial for agentic workflows, where AI systems can complete multi-step tasks autonomously.

Google is not releasing Argon to the general public first; instead, it's being rolled out through the Fairwind Program to 'trusted cyber defenders', a vetted group of security organizations and government partners. This decision is based on safety concerns, as a model this capable at identifying and exploiting vulnerabilities could be misused by bad actors.