Google announces Gemini 4 Argon AI model, but you can't use it yet

Key Points:

• Google has introduced Gemini 4 Argon, a new AI model claiming industry-leading performance in coding, knowledge work, and cybersecurity, though it is currently only accessible to internal engineers and select partners.
• Argon has been used internally to optimize memory usage across Google's data centers and assist in migrating large C/C++ codebases to Rust, including significant portions of core libraries and the Fuchsia OS kernel.
• Benchmark tests show Gemini 4 Argon outperforming competitors like GPT-6 Astra on software engineering and economic analysis tasks, while supporting a much higher output limit of 1 million tokens compared to previous models.
• The initial release focuses on cybersecurity, with trusted testers in Google's Fairwind Program using Argon to identify critical vulnerabilities missed by other frontier models, emphasizing the model's enhanced safety and monitoring systems.
• Google plans to eventually offer Gemini 4 Argon to enterprise and consumer customers, starting with paid API users and Google AI Ultra subscribers, but no specific timeline for general availability has been announced.