Safety Evaluation
Benchmarks
Dataset NameSOTA methodMetricTrendResultsLast Updated
48.3Illegal Activity Score
5
Feb 26, 2026
97.5Universal Human Values
5
Feb 26, 2026
79.2Refusal Rate
4
Jun 16, 2026
99.4Acceptance Rate
4
Jun 9, 2026
3.8Misalignment Rate
4
Jun 9, 2026
0Misalignment Rate
4
Jun 9, 2026
34.381-Clip Safety Score
4
Jun 2, 2026
48Availability Rate
4
Jun 2, 2026
81.7Safety Score
4
Jun 2, 2026
88.5Safe Rate
4
Jun 2, 2026
90Safe Rate
4
Jun 2, 2026
100Safe Rate
4
Jun 2, 2026
90Safe Rate
4
Jun 2, 2026
85Safe Rate
4
Jun 2, 2026
87Safety Score
4
Jun 2, 2026
100Utility Score
4
May 13, 2026
32.9Reasoning Accuracy (Avg)
4
May 7, 2026
85Overall Safety Score
4
Mar 18, 2026
69.9Safety Score
4
Mar 18, 2026
97.7Safety Score
4
Mar 18, 2026
0.001Cost per Accuracy Point ($)
4
Mar 6, 2026
0Cost per Accuracy Point ($)
4
Mar 6, 2026
0.001Cost per Accuracy Point ($)
4
Mar 6, 2026
0.002Cost per Accuracy Point ($)
4
Mar 6, 2026
100Safe Response Rate
4
Feb 26, 2026