Atlas

Benchmarks

← All benchmarks

Chrome Production Commit Scanning

Safety · 2026-07-21

Google’s private Chrome production commit-scanning pipeline evaluates whether a model identifies security issues in real production commits, reported as pass@1 without safety guardrails. The release does not disclose the commit set or its size, so Atlas retains the result as a diagnostic.

Top models (higher is better)

ModelScore
Gemini 3.5 Flash Cyber72.0
Gemini 3.5 Flash55.0
Opus 4.654.0
Loading Atlas data…