Skip to content
Trending storyDeveloping

Gemini 4 Argon scores 67.5% on depthfirst dfbench vulnerability detection

1 article1 sourcesince Oct 20Last article 1h ago ·

Overview

AISummary of 1 article

depthfirst reports that Gemini 4 Argon scored 67.5% across all 165 tasks on its dfbench benchmark for finding real vulnerabilities in code, second only to Mythos 5 at 69%.

On application and smart-contract tasks, where Gemini 4 Argon rarely refused, it found 74% of known vulnerabilities, which depthfirst calls the highest recall it has measured.

Google's Paige Bailey announced Gemini 4 Argon for security vulnerability detection at a lower price than competitors. Those price claims are Bailey's and are not covered by the benchmark results.

Written by AI from the articles below · updated Oct 20, 12:21 AM ET

Check the sources:

Article timeline

The articles in this story. Times are ET.

Oct 20
  1. 👩‍💻 Paige BaileyX
    Gemini 4 Argon finds 74% of known vulnerabilities on depthfirst benchmark

    AIGoogle's Paige Bailey announces Gemini 4 Argon for detecting security vulnerabilities at a lower price than competitors. In depthfirst's dfbench, Gemini 4 Argon found 74% of known vulnerabilities on application and smart-contract tasks, where it rarely refused. Across all 165 tasks it scored 67.5%, second only to Mythos 5 at 69%.

Heat Trend

Not enough continuous observations to show a trend yet.