Gemini 4 Argon: our next era of frontier intelligence

This post introduces Gemini 4 Argon, a new frontier model. It details its application in quantum algorithmic optimization (40% improvement), memory optimization across data centers (300+ TiB freed), and large-scale codebase migrations (C/C++ to Rust, including re2, libgav1, and Fuchsia Zircon kernel). For libgav1, Argon agents optimized a Rust port, achieving a 2.7x speedup over the Rust port with identical video output. It also sets new state-of-the-art on DeepSWE v1.1 (77.9%), Vals Index, Vals Finance Agent v2, Harvey's Legal Agent Benchmark, AutomationBench (51.3%), and LVBench (91.7%). In cybersecurity, it ties for first on CWE-bench v1 (68%) and demonstrates superior vulnerability discovery over 3.8 Flash Cyber on internal benchmarks and Wiz's black-box penetration testing benchmark. The post also outlines strengthened safeguards against misuse, prompt injection, and misalignment, and hardening of sandboxed environments.