The OWASP Benchmark is an official evaluation set that scores a security tool's accuracy across about 2,700 test cases as a Youden index from 0 to 100. Here is where Chitos lands, self-measured.

Configuration Youden

Base code scanner alone 52.8

Chitos, online judgment engine (Claude Opus 5.5) 77.2

Chitos, on-prem judgment engine (Darwin-180B-RSI) 77.8

Reading the jump

The gap from 52.8 to 77 is the AI verification layer. It removes more than 60 percent of false positives while missing none of the real vulnerabilities. A Youden index weights sensitivity and specificity together, so this is not a precision trick that quietly drops true positives; it is cleaner findings at the same recall.

Why the on-prem row matters most

The online engine and the on-prem engine scored effectively the same.…