WHAT NEW IN X11.05 THU

AI collections from X · 23 collections · 68,213 posts

gemini-4IMAGE

Gemini 4 Argon on dfbench: 74% vulnerability recall, 67.5% overall

Depthfirstlabs tested Gemini 4 Argon on its dfbench vulnerability benchmark: 74% recall on app and smart-contract tasks and 67.5% across 165 tasks, behind only Mythos 5.

Ram
@PothamRamArchived 16 days agoOriginally posted 2026.10.20
Gemini 4 Argon on dfbench: 74% vulnerability recall, 67.5% overall
Ram @PothamRamOriginally posted 2026.10.20
Open original on X

We tested Gemini 4 Argon on @depthfirstlabs dfbench, our benchmark for finding real vulnerabilities in code. On application and smart-contract tasks, where it rarely refused, it found 74% of known vulnerabilities, the highest recall we have measured. Across all 165 tasks it scored 67.5%, second only to Mythos 5 (69%{.

x.com/pothamram/status/2112309211854180666Archived 2026.10.20 09:59