THE DAILY DOWNLOAD / GOOGLE / GEMINI
Gemini 4 Argon hunts hackers. You can’t have it yet.
Google announced Gemini 4 Argon on Sept 30: its most powerful model, built for coding, research and cyber defence. Security pros get it first; everyone else waits.
The TL;DR
- Gemini 4 Argon is Google's new top model for real-world coding, office knowledge work and cyber defence.
- It can write up to 1 million tokens in one answer: roughly several novels' worth of text.
- Only vetted cyber defenders have it right now. Google AI Ultra subscribers and paying developers are next, with no date yet.
Google said "my turn" on Sept 30 with Gemini 4 Argon, its new top-tier model. The pitch: a model that codes, researches, crunches company documents and, most notably, helps defend software by finding and fixing security holes.
The stats are giving overachiever. Google says Argon scored 77.9% on DeepSWE v1.1, a test of real software engineering, and ranked first on AutomationBench with 51.3%. It can also write up to 1 million tokens in one go, up from 64,000. Usual reminder: these are Google's own numbers, so wait for independent tests before you crown it.
The catch: you can't use it yet. Because it's so good at security work, Google is rolling it out first to trusted cyber defenders through its Fairwind Program. Google AI Ultra subscribers, its priciest consumer plan, and paying developers come next "as soon as possible."
Developers will pay an introductory $2 per million input tokens and $10 per million output, rising later to $4 and $20.
Why you should care
AI that can find security holes could also be used to exploit them, so a slow, vetted rollout is the responsible move. Nothing changes in the Gemini app today, but tech like this could end up quietly guarding the software you use every day.
Vibe check
Big aura. Velvet rope.