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

#ModelScoreOverall rank
1GLM-5.171.1#195
2GPT-571#177
3GPT-5.565.2#50
4DeepSeek V4 Pro62.5#198
5Kimi K2.661.5#191
6GPT-5.4 Mini37.4#401
7DeepSeek V4 Flash30.5#265
8Gemma 4 31B30.5#385
9Qwen 3.6 35B A3B26.1#484
10Grok 4.1 Fast19.2#429
11GPT-OSS-120B9.5#707
12Gemma 4 26B A4B4.8#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.