RevengeBench - RobotRumble: leaderboard
RevengeBench asks an agent to reconstruct a hidden game bot's policy from observed behavior and experiments. This series reports RobotRumble distance reduction relative to a starter policy, in percent, under mini-SWE-agent with the best of five rounds reported. It measures how closely the recovered policy matches held-out actions; higher is better and negative values mean worse than the starter.
Metric: Distance reduction (%; best round, mini-SWE-agent). Source: revengebench.com. Saturation forecast: Around May 2027. 12 models tracked. Updated 2 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GLM-5.1 | 71.1 | #195 |
| 2 | GPT-5 | 71 | #177 |
| 3 | GPT-5.5 | 65.2 | #50 |
| 4 | DeepSeek V4 Pro | 62.5 | #198 |
| 5 | Kimi K2.6 | 61.5 | #191 |
| 6 | GPT-5.4 Mini | 37.4 | #401 |
| 7 | DeepSeek V4 Flash | 30.5 | #265 |
| 8 | Gemma 4 31B | 30.5 | #385 |
| 9 | Qwen 3.6 35B A3B | 26.1 | #484 |
| 10 | Grok 4.1 Fast | 19.2 | #429 |
| 11 | GPT-OSS-120B | 9.5 | #707 |
| 12 | Gemma 4 26B A4B | 4.8 | #581 |
No result here: #1 Claude Fable 5.2, #2 GPT-6.2 Astra, #3 Gemini 4 Carbon.
Interactive version: theaggregate.ai/benchmark?slug=revengebench-robotrumble · How It Works · Data refreshed daily, snapshot 2026-11-17.