Atlas

Benchmarks

← All benchmarks

CAISI Biology Safeguards - No Jailbreak

Safety · 2026-07-17

The no-jailbreak condition of CAISI Biology Safeguards reports the percentage of responses graded fully compliant with sensitive biological requests. The number of private benchmark items was not disclosed.

Top models (lower is better)

ModelScore
Opus 4.80.0
GPT-5.50.0
GLM-5.214.0
Loading Atlas data…