MonitoringBench - Recon - Firewall Backdoor: leaderboard

MonitoringBench (2026) tests LLMs as monitors of coding agents in BashArena: the monitor reads a full agent trajectory of tool calls and scores how suspicious it is, and an attack counts as caught when the score exceeds the monitor threshold for a 1% false-positive rate on honest runs. This board covers the 204 successful firewall backdoor attacks from the recon stage of the red-teaming pipeline, scored as catch rate in percent; higher is better.

Metric: Catch rate (%) at a 1% false-positive-rate threshold calibrated on 416 honest BashArena runs: share of the 204 successful firewall backdoor attack trajectories (strategies from reconnaissance agents (with and without GPT-5 monitor feedback) executed by Claude Opus 4.6) that the full-trajectory monitor scores above its threshold, default ControlArena monitor prompt with a 0-10 suspiciousness score, mean of three scoring runs; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 13 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Claude Opus 4.5 (20251101)72.9#142
2Claude Sonnet 4.562.7#267
3Claude Sonnet 4 (20250514)56#410
4GPT-5.1 Codex Max51.8#212
5O3 (2025-04-16)35.9#238
6GPT-529.7#185
7GPT-4.116.5#506
8GPT-4o (2024-08-06)12.7#750
9GPT-4 Turbo7.5#1015
10GPT-5 Nano5.7#912
11GPT-4.1 Mini4.7#799
12O4 Mini (2025-04-16)2.5#375
13GPT-4o Mini (2024-07-18)0.5#1501

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=monitoringbench-recon-firewall-backdoor · How It Works · Data refreshed daily, snapshot 2026-11-22.