[Analysis: Google Launches Gemini 4 to Catch Up with AI Rivals, but Internal Opinions on Its Programming Performance Are Divided] Google has begun rolling out its highly anticipated flagship AI model, Gemini 4 Argon, but internal doubts remain about the model's real-world performance in critical areas such as programming. On Wednesday, Google made the model available to a small group of trusted cybersecurity partners, stating that it plans to expand its usage after further testing, starting with paid users. Gemini 4 achieved leading results in several benchmark tests, including outperforming OpenAI's Astra model in a test measuring security capabilities. However, some insiders argue that these metrics do not tell the whole story. According to sources directly familiar with the project, while Gemini 4 excelled in benchmark tests designed to measure model performance, its real-world effectiveness, particularly in certain programming tasks, has been less impressive during internal use by employees.