SAI LabsBook a demo

Safety IndexEvaluationsAIR-Bench hate and harassment safe rate (lab-reported)

AIR-Bench hate and harassment safe rate (lab-reported)

Refusal calibration · Harmful requestsDeveloper-reported

The developer's own single-turn hate-and-harassment safe-response rate.

Metric
% safe. Higher is better.
Dataset
arxiv.org/abs/2407.17436
Grader
the developer's own grader (see the model's system card)
Status
–. Measured on 1 model, 0 readings by SAI.
Notes
From Google's Gemini 4 Argon Safety & Security Report; internal data, autorater unstated. Developer-reported; never counted.

Results

ModelValueScoreSourceSetting
Gemini 4 Argon92.7%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 configurationsingle-turn safe response rate