Yooo, Google just unveiled Gemini 4 Argon, and this isn't being introduced as just another AI chatbot. Google built this model to stay on complicated jobs for much longer, including software engineering, financial research, legal work and cybersecurity.
THIS AI CAN ACTUALLY FIND AND FIX SECURITY FLAWS
Google says Argon can autonomously find, validate AND patch critical software vulnerabilities. That's a big difference from simply asking AI to look at some code and tell you what's wrong.
Cybersecurity company Wiz is already using Argon through its Scan for Good initiative. According to Google, Argon discovered a critical vulnerability affecting healthcare software used by hospitals around the world that previous frontier AI models had missed.
1 MILLION OUTPUT TOKENS ?
Google also increased Argon's maximum output from 64,000 tokens to 1 MILLION tokens. Basically, Google is giving the model a lot more room to keep working through huge, complicated problems without constantly having to stop.
Argon scored 77.9% on DeepSWE v1.1, which tests long-horizon software engineering. It also scored 68% on CWE-bench v1, tying for the top score on that security vulnerability remediation benchmark.
GOOGLE IS ALREADY USING ARGON INTERNALLY
Thousands of Google employees are already using Argon. Google says Argon agents have helped with projects ranging from migrating massive C/C++ codebases to Rust to finding memory optimizations across Google's data centers.
One of those projects freed more than 300 TiB of memory after deployment. Google also says Argon is being used on codebase migrations reaching more than 800,000 lines of code.
BUT YOU CAN'T JUST OPEN GEMINI AND USE ARGON YET
An AI capable of autonomously discovering software vulnerabilities can obviously become dangerous in the wrong hands. That's why Google is starting Argon's rollout with trusted cybersecurity defenders through its Fairwind Program while it continues testing safeguards.
Google says broader access for developers, enterprises and consumers is coming. Introductory API pricing will start at $2 per million input tokens and $10 per million output tokens.
This is where AI is heading. We're moving past simply asking AI a question and getting closer to giving it a complicated problem and saying, “Work on this until it's solved.”
That's a HUGE difference.
Source: Google — Gemini 4 Argon
— King Coop
Content & Social Media Manager
