Google's Gemini 4 Argon tops rivals on key benchmarks, but cyber defenders get it before anyone else.
Google announced Gemini 4 Argon on September 30 as its new frontier model, built for deep reasoning on long, complex tasks.
Argon can write up to 1 million output tokens, up from 64,000. Google says it can find, validate and patch software vulnerabilities.
On DeepSWE v1.1, Argon scored 77.9% versus 74.2% for Claude Opus 5.5 and 74.1% for GPT-6 Astra, per Google's own figures.
Trusted cyber defenders in the Fairwind Program get it now. Paid API customers and Google AI Ultra subscribers are next, with no date yet.
Nothing changes in the Gemini app today. Watch for the AI Ultra rollout as the sign Google thinks Argon is ready for regular people.
Google’s Gemini 4 Argon Is Here, but Your Phone Won’t Get It Yet