Google Unveils Gemini 4 Argon with Focus on Coding and Cyber Defense
Google’s latest Gemini model introduces advanced capabilities for handling complex, long-running tasks, with a particular focus on software development and cybersecurity. Gemini 4 Argon will initially be available to selected cyber defense experts through the Fairwind Program, before access is gradually expanded to developers, enterprises, and other users. The model can autonomously identify, validate, and patch critical software vulnerabilities, while Wiz is already using it as part of its Scan for Good initiative, where, according to Google, Argon discovered a critical vulnerability in healthcare software used by hospitals worldwide.
Argon has also been designed for demanding tasks in software engineering, finance, and legal workflows, and Google says thousands of its employees are already using the model internally. A team of AI agents powered by Argon analyzed telemetry data from Google’s data centers and identified optimizations that freed up more than 300 TiB of memory after deployment. To support more complex and longer-running tasks, the model’s maximum output has been increased from 64,000 to one million tokens, while Argon achieved a score of 77.9 percent on the DeepSWE v1.1 benchmark.
Introductory pricing will be $2 per million input tokens and $10 per million output tokens, with cached input tokens priced 95 percent lower. Before making the model more widely available, Google plans to further strengthen safeguards against misuse, prompt injection attacks, and other security risks, after which Argon is expected to gradually become available to paid API users, enterprises, and Google AI Ultra subscribers.
