🛡️Google Gemini 4 Argon Rolls Out to Cyber Defenders With 1M Token Limit

Your security team just got a 1M-token coding agent

Google just put its new frontier model, Gemini 4 Argon, in the hands of trusted cyber defenders first. Public access comes later, after safety testing. The headline spec: output tokens jump from 64K to 1M, and the input window is 1M too. If you do security work, this is the one to watch. Argon is being used to autonomously find, validate, and patch critical vulnerabilities, and it beat the 3.8 Flash Cyber model at discovering attack surfaces and producing proof-of-concept evidence. If you're running pentests, that's a workflow change. The same model is also being used for large C/C++ to Rust migrations, so platform teams should pay attention too. The numbers Google has shared are concrete. Argon's memory optimization work across its data centers freed more than 300 TiB of memory. Vulnerability discovery covers codebases in 20 programming languages. Google says the model refuses harmful requests while still supporting dual-use science, under its Frontier Safety Framework. Watch the public release timeline. Google hasn't given a date, and the safety gating suggests the broader rollout will be slower than the internal use suggests.