Wiz researchers say Gemini 4 Argon found over twice as many vulnerabilities as baseline models

In a Google video, Wiz researchers said Gemini 4 Argon found "over two times more vulnerabilities than other baseline models" in "shadow A/B testing on real world systems," "while using less tokens." Google's post on X repeats the claim, saying Argon found more than twice as many vulnerabilities as baseline models and caught a critical risk that previous frontier models missed.

What hasn't been disclosed. Neither Wiz nor Google named the baseline models or published vulnerability counts, testing methods or data, so the 2x figure from Wiz researchers and Google can't be independently checked.

Published benchmarks. Google's Sep 30 launch blog says Argon outperforms Gemini 3.8 Flash Cyber on Wiz's internal black-box penetration-testing benchmark, without giving numbers. It ties for first at 68% on CWE-bench v1, which measures fixing flaws, not finding them.

A healthcare case. According to Google DeepMind, Argon found a critical flaw in hospital software used worldwide that previous frontier models had missed. A bypass in a file-download feature gave access to the app's master key and admin rights. DeepMind says finding it proactively safeguarded hundreds of thousands of patient records. The vendor was not named.

Availability. Gemini 4 Argon has been generally available since Oct 13 through the API (AI Studio and the Gemini platform), the Gemini app, AI Mode in Search, and Antigravity for Ultra subscribers. A version without cyber guardrails is limited to trusted defenders in the Fairwind Program and Google's internal teams.

Sources: Google video, Google on X, Google blog, Google DeepMind