RevengeBench - BattleSnake: leaderboard

RevengeBench asks an agent to reconstruct a hidden game bot's policy from observed behavior and experiments. This series reports BattleSnake 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 June 2027. 12 models tracked. Updated 2 Oct 2026.

Top models

#ModelScoreOverall rank
1GPT-577.6#177
2GPT-5.573.5#50
3DeepSeek V4 Pro72.3#198
4Kimi K2.670.6#191
5GLM-5.170.5#195
6Gemma 4 31B66.4#385
7Gemma 4 26B A4B64.9#581
8DeepSeek V4 Flash60.9#265
9GPT-5.4 Mini51.5#401
10Qwen 3.6 35B A3B48.9#484
11Grok 4.1 Fast41.3#429
12GPT-OSS-120B40.8#707

No result here: #1 Claude Fable 5.2, #2 GPT-6.2 Astra, #3 Gemini 4 Carbon.

Interactive version: theaggregate.ai/benchmark?slug=revengebench-battlesnake · How It Works · Data refreshed daily, snapshot 2026-11-17.