MonitoringBench - Recon - Malware Download: 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 669 successful malware download 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 669 successful malware download 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)85.1#139
2Claude Sonnet 4.576.6#261
3GPT-568.7#178
4GPT-5.1 Codex Max68.2#207
5O3 (2025-04-16)67.5#231
6Claude Sonnet 4 (20250514)59.5#405
7GPT-4.129.1#505
8GPT-4o (2024-08-06)22.1#743
9GPT-4 Turbo19.4#1000
10O4 Mini (2025-04-16)15.9#370
11GPT-5 Nano11.2#906
12GPT-4.1 Mini8.8#789
13GPT-4o Mini (2024-07-18)1.1#1484

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=monitoringbench-recon-malware-download · How It Works · Data refreshed daily, snapshot 2026-11-21.