SAI LabsBook a demo

Safety IndexEvaluationsStrongREJECT (lab-reported)

StrongREJECT (lab-reported)

Refusal calibration · Harmful requestsDeveloper-reported

The developer's own StrongREJECT-style result; each lab words the metric differently, and the wording is in the setting column.

Metric
% safe / not unsafe. Higher is better.
Dataset
arxiv.org/abs/2402.10260
Grader
the developer's own grader (see the model's system card)
Status
–. Measured on 4 models, 0 readings by SAI.
Notes
Lab system cards; OpenAI reports not_unsafe on its adaptation with known jailbreaks inserted; Google reports a safe-response rate. Developer-reported; never counted.

Results

ModelValueScoreSourceSetting
Gemini 4 Argon96.6%developer-reportedGoogle, Gemini 4 Argon Safety & Security Report
Developer-reported · 2026-10
Reported by the model's developer in its own system card or report.
Full configurationSafe Response Rate (%), single-turn, higher is better; StrongREJECT 313 no jailbreak
GPT-597.4%developer-reportedOpenAI (GPT-5.1 system card addendum)
Developer-reported · 2025-11-12
Reported by the model's developer in its own system card or report.
Full configurationnot_unsafe, aggregate (OpenAI StrongREJECT adaptation, known jailbreaks inserted); Table 2; higher is better
GPT-5.196.7%developer-reportedOpenAI (GPT-5.1 system card addendum)
Developer-reported · 2025-11-12
Reported by the model's developer in its own system card or report.
Full configurationnot_unsafe, aggregate; Table 2; higher is better
GPT-5.297.5%developer-reportedOpenAI (GPT-5.2 system card update)
Developer-reported · 2025-12-11
Reported by the model's developer in its own system card or report.
Full configurationnot_unsafe, aggregate, filtered subset (saturated examples removed); Table 2; higher is better