RevengeBench - Halite: leaderboard

Metric: Distance reduction (%; best round, mini-SWE-agent). Source: revengebench.com. Saturation forecast: Around September 2027. 12 models tracked.

Top models

#ModelScore
1GPT-5.4 Mini73.3
2Grok 4.1 Fast66.7
3Kimi K2.664.4
4GPT-5.563.9
5Gemma 4 26B A4B62.5
6Gemma 4 31B62.3
7GPT-561.4
8GLM-5.160.8
9GPT-OSS-120B59.8
10DeepSeek V4 Pro42.6
11Qwen 3.6 35B A3B41.5
12DeepSeek V4 Flash37.1

Interactive version: theaggregate.ai/benchmark?slug=revengebench-halite · How It Works · Data refreshed daily, snapshot 2026-10-02.