Gemini 4 Argon: our next era of frontier intelligence

Google DeepMind and Chief AI Architect, Google have announced the launch of their new frontier model, Gemini 4 Argon. This model is being rolled out to a select group of cyber defenders as part of the company's Fairwind Program. Gemini 4 Argon is designed to handle complex, long-horizon workflows, delivering frontier performance in areas such as software engineering, enterprise knowledge work, and cybersecurity defense.

The model is currently being tested by a group of trusted cyber defenders and will undergo further refinement before being made available to developers, enterprises, and consumers. Google engineers have already found value in Gemini 4 Argon, utilizing it for tasks ranging from debugging to large-scale codebase migrations and algorithm designs. The model has set a new state of the art in DeepSWE v1.1, a metric that measures a model's performance in real-world long-horizon software engineering tasks.

In addition to its software engineering capabilities, Gemini 4 Argon has excelled in other domains. It ranks #1 on AutomationBench, a Zapier benchmark measuring end-to-end execution across core business functions. The model also demonstrates strong performance in long video understanding, ranking state of the art on LVBench with a score of 91.7%. Furthermore, Gemini 4 Argon has shown impressive vulnerability discovery capabilities, tying for first place on CWE-bench v1 with a top score of 68%.

To support the model's advanced capabilities, Google is significantly expanding the output token limit to an industry-leading 1 million tokens, up from the previous 64K tokens. This allows the model to sustain deep reasoning and generate hundreds of thousands of tokens in a single trajectory.