OpenCompass Reasoning - Academic — leaderboard

Metric: Score (%). Source: rank.opencompass.org.cn. 23 models tracked.

Top models

#ModelScore
1GPT-5.4 (High)52
2Claude Opus 4.7 (High)50.9
3Kimi K2.648.8
4Claude Sonnet 4.6 (High)45.6
5DeepSeek V4 Pro43.9
6Gemini 3.1 Pro (Preview)43.6
7Hy3-preview (High)43.6
8Qwen 3.6 Max Preview43.3
9GPT-5.4 Mini (High)41.6
10DeepSeek V4 Flash39
11GLM-5.138.7
12Step 3.5 Flash38.7
13Gemma 4 31B (IT)38.1
14Qwen 3.5 397B A17B36.6
15Qwen 3.6 27B36

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