Google Unveils Gemini 4 Argon AI Model with Enhanced Capabilities

Google has released its latest AI model, Gemini 4 Argon, which boasts significant improvements in complex tasks and workflows. The new model can handle longer and more complicated problems without interruption, thanks to a greatly increased output token limit of 1 million tokens.

This is a major upgrade from the previous limit of 64,000 tokens, allowing Argon to sustain deeper reasoning and produce hundreds of thousands of tokens during a single workflow. Google employees have already been testing the model for various tasks, including coding, research, writing, and cybersecurity defense.

The company claims that Argon has helped its researchers optimize quantum computing algorithms, improving one example by 40% within minutes. Additionally, Argon agents are being used to analyze memory usage across data centers, resulting in significant savings of over 300 TiB of memory, with potential total savings of up to 1 PiB.

When wider access begins, Gemini 4 Argon will have an introductory price of $2 per million input tokens and $10 per million output tokens, with cached input tokens offered at a 95% discount compared to the standard input price. Google is initially rolling out the model to trusted cybersecurity defenders through its Fairwind Program while continuing testing and improving safeguards.