Google has announced Gemini 4 Argon, the most powerful artificial intelligence model in its history. The network's main technological breakthrough is an expanded output limit of 1 million tokens per request, far surpassing previous generations and enabling large-scale tasks to be completed in a single reasoning cycle, according to infohub.kz.
The new flagship AI model was built to handle complex tasks in programming, finance and cybersecurity, TechCrunch reports.
Gemini 4 Argon posted record results on specialized tests for code writing, financial analysis, drafting legal documents and analyzing long video materials.
The developers placed particular emphasis on cybersecurity: the model can autonomously find vulnerabilities in software and generate fixes to close the gaps.
Inside Google, Argon-based agents are already being used to automatically rewrite code from C and C++ into the safer Rust language, as well as to optimize memory usage in data centers.
Initially, Google is granting access to the model to a limited group of vetted information security specialists under the Fairwind program, as well as to developers within the company.
A full-scale release and public access to the API interface are planned after a series of security tests is completed.