Google has announced a new artificial intelligence model, Gemini 4 Argon, which is already being actively used in its internal processes. This model is designed to perform complex, multi-step tasks such as programming, working with corporate information, and cybersecurity.
Gemini 4 Argon is capable of handling long-running workflows that require processing large volumes of data. In particular, the input data capacity has been increased to 1 million tokens, allowing the model to generate hundreds of thousands of tokens in a single session. This significantly improves the efficiency of performing complex tasks.
Argon’s primary applications include specialized programming, in-depth research, text generation, and algorithm development. For example, the model assists in porting codebases from C and C++ to Rust, which involves processing tens of thousands of lines of code.
Google also uses Argon to optimize quantum computing algorithms, where the model demonstrates significant improvements in resource conservation. It has identified opportunities to reduce memory consumption in data centers, freeing up over 300 TB of memory.
In the field of cybersecurity, Gemini 4 Argon is capable of detecting and fixing critical software vulnerabilities. The model is already being used by Wiz to protect critical infrastructure, identifying vulnerabilities that were not detected by previous versions of the model.
Google currently plans to gradually expand access to Gemini 4 Argon as part of a voluntary process with the U.S. government. The company is also strengthening security measures to prevent potential misuse of the technology, particularly in the context of cyberattacks.
