Reward Hacking Benchmark (RHB) — leaderboard

Agentic reward-hacking benchmark with multi-step tool-use tasks containing naturalistic shortcut opportunities such as skipping verification, exploiting metadata, or tampering with evaluation-relevant functions.

Metric: Exploit rate (%) (self-reported). Source: benchmarklist.com. Status: saturation imminent. 13 models tracked.

Top models

#ModelScore
1O311.8
2O4 Mini8.4
3O3 Mini7.1
4O16.8
5Gemini 2.5 Pro (Preview 06-05)4.6
6Claude 3.7 Sonnet3.9
7GPT-4o0.9
8Gemini 2.5 Flash (Preview 09-2025)0.8
9Claude 3.5 Sonnet0.6
10DeepSeek V30.6
11Claude Sonnet 4.50
12Claude Opus 4.50

Interactive version: theaggregate.ai/benchmark?slug=reward-hacking-benchmark-rhb · How the rankings work · Data refreshed daily, snapshot 2026-07-22.