Google announces Gemini 4 Argon AI model, but it remains unavailable to the public

Google has officially unveiled its latest frontier AI model, Gemini 4 Argon, positioning it as a leader in coding, cybersecurity, and complex knowledge work. Despite the announcement, the model is currently restricted to internal use by Google engineers. According to the company, Argon has already demonstrated significant utility, including optimizing data center memory usage by 300 TiB and automating the migration of massive C/C++ codebases to Rust, notably within the Fuchsia OS Zircon kernel. Google supports these performance claims with benchmarks such as DeepSWE v1.1, where Argon achieved a 77.9 percent score, reportedly outperforming competitors like GPT-6 Astra and Opus 5.5. While the company highlights Argon's success in long-horizon tasks and economic analysis, it has not provided a specific timeline for public access. For now, the model serves as a powerful internal tool for Google's infrastructure and development teams.