PreAct-Bench - Cybersecurity: leaderboard

Metric: Prefix Foresight F1 with the full prefix (PFF(100), binary F1 of the unethical class, 0-1 scaled to %): given an agent trajectory truncated just before its first overt unethical step, the monitor predicts whether the trajectory will turn unethical; paired ethical and unethical trajectories per task; default generation parameters; higher is better; the cybersecurity domain. Source: arxiv.org. Saturation forecast: Around November 2027. 18 models tracked.

Top models

#ModelScore
1Gemma 3 27B89
2GPT-5.485.5
3Qwen 3 30B A3B 2507 Instruct84.8
4Qwen 3 8B (Non-reasoning)76.2
5Gemma 3 4B66.7
6Qwen 3 0.6B (Non-reasoning)66.7
7Qwen 3 8B (Thinking)63.7
8Qwen 3 0.6B (Thinking)62.6
9Qwen 3 30B A3B 2507 (Thinking)60.7
10GPT-OSS-20B59.6
11GPT-4o51.9
12Qwen 3 4B 2507 Instruct51.9
13GPT-OSS-120B48.9
14Qwen 3 4B 2507 (Thinking)43

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