MonitoringBench — leaderboard
Metric: Baseline catch rate (%) (self-reported). Source: benchmarklist.com. 13 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 4.5 | 95.1 |
| 2 | Claude Sonnet 4.5 | 83.5 |
| 3 | GPT-5.1 Codex | 71.1 |
| 4 | O3 | 64.4 |
| 5 | GPT-5 | 62.7 |
| 6 | Claude Sonnet 4 | 58.7 |
| 7 | GPT-4.1 | 11 |
| 8 | GPT-4o | 7.6 |
| 9 | GPT-5 Nano | 5.3 |
| 10 | O4 Mini | 3.2 |
| 11 | GPT-4 Turbo | 2.5 |
| 12 | GPT-4.1 Mini | 0.5 |
| 13 | GPT-4o Mini | 0 |
Interactive version: theaggregate.ai/benchmark?slug=monitoringbench · How the rankings work · Data refreshed daily, snapshot 2026-07-22.