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

Google released the Gemini 4 Argon model, claiming industry-leading performance in coding, knowledge work, and cybersecurity. Argon is used internally to process fleet-wide telemetry, saving 300 TiB of memory, and to migrate thousands of C/C++ lines to Rust in libraries such as re2, libgav1, and the Fuchsia OS Zircon kernel. Benchmark results show Argon scoring 77.9% on DeepSWE v1.1, surpassing GPT-6 Astra, Fable 5.1, and Opus 5.5, and achieving top marks on the Vals Index economic analysis test.