KORGym - Puzzle — leaderboard

Metric: Score. Source: razor233.github.io. 19 models tracked.

Top models

#ModelScore
1Gemini 2.5 Pro (03-25)0.94
2Claude 3.7 Sonnet (Thinking)0.93
3O1 (2024-12-17)0.83
4DeepSeek R10.82
5O3 Mini0.81
6DeepSeek V3 (0324)0.55
7Qwen 3 32B (Thinking)0.55
8Gemini 2.0 Flash (Thinking)0.53
9Gemini 2.0 Flash0.28
10DeepSeek R1 Distill Qwen 32B0.28
11GPT-4o0.25
12Doubao-1.5-Pro0.16
13Qwen 2.5 72B Instruct0.1
14Qwen 2.5 32B Instruct0.07
15DeepSeek-R1-Distill-Qwen-7B0.02

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