MonitoringBench: leaderboard
We introduce a red-teaming methodology that exposes harder-to-catch attacks for coding-agent monitors, suggesting that current practices may under-elicit attacks and overstate monitor performance. We identify three challenges with current red-teaming. First, mode collapse in attack generation, which we reduce with a novel attack taxonomy for broader coverage. Second, a conceive-execute gap: fronti
Metric: Baseline catch rate (%) (self-reported). Source: benchmarklist.com. Saturation forecast: Around January 2027. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.5 | 95.1 | #83 |
| 2 | Claude Sonnet 4.5 | 83.5 | #142 |
| 3 | GPT-5.1 Codex | 71.1 | #159 |
| 4 | O3 | 64.4 | #124 |
| 5 | GPT-5 | 62.7 | #98 |
| 6 | Claude Sonnet 4 | 58.7 | #199 |
| 7 | GPT-4.1 | 11 | #248 |
| 8 | GPT-4o | 7.6 | #343 |
| 9 | GPT-5 Nano | 5.3 | #425 |
| 10 | O4 Mini | 3.2 | #180 |
| 11 | GPT-4 Turbo | 2.5 | #404 |
| 12 | GPT-4.1 Mini | 0.5 | #355 |
| 13 | GPT-4o Mini | 0 | #599 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=monitoringbench · How It Works · Data refreshed daily, snapshot 2026-10-20.