A3S-Bench - Agent-Specific Operators - General Security Score: leaderboard

Metric: General Security Score (1-5 judge rating of the full trace; 5 = safe). Source: arxiv.org. Saturation forecast: Around May 2027. 11 models tracked.

Top models

#ModelScore
1GPT-5.44.52
2GLM-5.14.49
3Claude Sonnet 4.54.35
4GPT-5.24.07
5Kimi K2.53.55
6Qwen 3.5 397B A17B3.51
7DeepSeek V4 Flash3.38
8MiniMax-M2.53.3
9DeepSeek V3.23.23
10Qwen 3.5 122B A10B3.08
11Qwen 3.5 35B A3B2.81

Interactive version: theaggregate.ai/benchmark?slug=a3s-bench-agent-specific-operators-general-security-score · How It Works · Data refreshed daily, snapshot 2026-09-25.