OpenCompass Language - Creation — leaderboard

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

Top models

#ModelScore
1GPT-5.4 (High)82.5
2Qwen 3.6 Max Preview77.9
3Hy3-preview (High)75.4
4Gemini 3.1 Pro (Preview)73.8
5Kimi K2.673.8
6GPT-5.4 Mini (High)72.1
7Claude Sonnet 4.6 (High)70.8
8DeepSeek V4 Pro70
9GLM-5.169.2
10Claude Opus 4.7 (High)67.1
11MiniMax-M2.766.7
12Qwen 3.5 397B A17B66.2
13Qwen 3.6 27B65.8
14DeepSeek V4 Flash65
15Step 3.5 Flash63.8

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