SafePyramid: leaderboard

Metric: Rule matching rate (%) over SafePyramid's 3,000 cases (1,000 conversations in 10 domains, each with an L0, L1 and L2 policy): per case (a multi-turn conversation plus an in-context safety policy), the predicted set of violated rules matches the ground-truth set when neither false positives nor false negatives exceed (1 - tau) of the true violations; averaged over tau = 0.7, 0.8, 0.9 and 1.0; per-policy evaluation, refused cases excluded; higher is better. Source: arxiv.org. Saturation forecast: Around April 2028. 12 models tracked.

Top models

#ModelScore
1GPT-5.5 (xHigh)54.2
2Claude Opus 4.7 (Max)52.9
3DeepSeek V4 Pro (Max)52.2
4Seed 2.0 Pro (High)50
5Gemini 3.5 Flash (High)49.6
6Hy3-preview (High)49.1
7Qwen 3.6 Max (Preview) (Thinking)47.4
8Grok 4.3 (High)40.6
9gpt-oss-safeguard-20B22.5

Interactive version: theaggregate.ai/benchmark?slug=safepyramid · How It Works · Data refreshed daily, snapshot 2026-09-29.