Google releases Gemini 4 Argon: Chrome's six-million-line legacy refactor drops from months to weeks

Google positions Gemini 4 Argon as a model for complex real-world tasks across coding, finance, legal, cybersecurity, and science, and says it ranks first on the Vals Index, which measures performance across economic sectors such as finance, coding, legal, and tax work. Previous frontier models had limited depth and precision on these multi-step professional tasks, and Google says Argon outperforms other frontier models on benchmarks across domains. The comparison Google gives is the number-one Vals Index ranking, plus internal and early-access enterprise use across coding, legal, finance, and science workflows.

At Chrome, Argon helped separate core webpage handling from a roughly six-million-line legacy folder of browser features inside Chromium's roughly 40-million-line codebase, piping multiple clean code changes at a time and finishing an architectural refactor across hundreds of files. That work was expected to take months, and Google says the team completed it in weeks. Google says the payoff for Chrome's billions of global users is that live webpages can move between tabs, side panels, and floating windows without reloading, eliminating an entire category of browser crashes.

Argon in Antigravity helped Google software engineers increase agentic code submissions by more than 35% while reducing code rollbacks by nearly a third. The prior submission and rollback rates are the comparison baseline for that change. Google says this is what it observed over the past few weeks as internal software engineers used Argon.

On Google's open-source Agent Development Kit Python codebase, Argon worked autonomously overnight and surfaced optimizations that cut in-memory session prep time by 40% in long sessions and sped up cold imports by 21%. The team gave Argon only the .agents/skills rulebook and a one-paragraph goal to eliminate framework overhead without breaking existing apps, with no step-by-step guidance. Google says the optimizations are now live in google/adk-python.

This report covers Google's release description of Gemini 4 Argon and its internal and early-access cases, for researchers and engineers tracking how frontier models land in engineering and scientific workflows. Google says the model is available today through the Gemini API, AI Studio, Google Antigravity, and the Gemini app and AI Mode in Search for developers, enterprises, and subscribers, and it opened applications for 10,000 scientists to get Antigravity's highest usage tier free for six months, plus Fairwind Program applications for cybersecurity defenders.

Google says Argon ranks first on the Vals Index and scores 87% on ERQA, but those comparisons come from Google's own evaluations and internal tests. Most engineering and science cases are internal or early-access partner reports; some results are backed by public repository commits or laboratory testing, and the rest are vendor-reported. On safety, Google says external bio red teaming could unlock less than 1% of bio sub-tasks, violations on an adversarial set reduced nearly two-fold, shortcut attempts dropped to 2%, and attack success on the Gray Swan indirect prompt injection benchmark fell to 0.7% across 15 adaptive attempts.