AI agents have become a major headache for OpenAI. Recently, they made headlines due to problematic behavior. Now, the company is spending approximately ₹5 crore daily on investigations to address the issues caused by these AI agents.
Google has launched Gemini 4 Argon, designed for complex coding, research, enterprise work, and cybersecurity tasks. Currently, access to it is limited to select trusted cyber defenders.
Tech company Google has launched its new frontier AI model, Gemini 4 Argon. According to the company, it is its most powerful AI model to date, specifically designed to handle complex and long-duration tasks. Gemini 4 Argon can be utilized in fields such as coding, software engineering, professional work, research, and cybersecurity. However, it is not currently available to general users, and initial access remains limited.
The standout feature of Gemini 4 Argon is its 1-million output token limit. According to Google, this represents a significant upgrade from the previous 64,000-token capacity. Such a massive output capacity enables the model to handle long, multi-step tasks over extended periods and generate large volumes of output in a single process. This is particularly beneficial for professional tasks involving extensive codebases, in-depth research, and work in the legal and financial sectors.
Being Used for Such Tasks
According to the company, Gemini 4 Argon is also being utilized for real-world software engineering tasks. Google engineers are employing it for activities such as debugging, algorithm design, and large-scale codebase migrations. Google reported that the model achieved a score of 77.9 percent on DeepSWE v1.1. The company states that its agents are being used for major tasks like migrating C/C++ codebases to Rust—including codebases exceeding 800,000 lines of code.
Focus on Cybersecurity
Cybersecurity is also a key focus area for Gemini 4 Argon. According to Google, the model is capable of identifying software vulnerabilities, validating them, and patching critical flaws. Google’s cybersecurity partner, Wiz, has utilized Argon as part of its “Scan for Good” initiative. The company reported that, during the initial phase, the model discovered a critical flaw in healthcare software that posed a risk of leaking sensitive personal information.
