OpenCompass Subjective - Creation - English (CompassBench 2405): leaderboard

Metric: Score (%). Source: rank.opencompass.org.cn. 39 models tracked.

Top models

#ModelScore
1GPT-4o (2024-05-13)66.6
2GPT-4 Turbo60.1
3DeepSeek V2 Chat51.1
4Claude 3 Opus46.8
5Yi Large45
6Llama 3 70B Instruct44.7
7GLM-443.7
8Moonshot v1 8K41.6
9Mixtral 8x22B Instruct (v0.1)41.2
10Mistral Large41
11Command-R+40.6
12Llama 3 8B Instruct37.3
13Qwen 1.5 110B Chat37.3
14Mixtral 8x7B Instruct (v0.1)36.5
15Mistral 7B Instruct (v0.3)34.1

Interactive version: theaggregate.ai/benchmark?slug=opencompass-subjective-creation-english-compassbench-2405 · How It Works · Data refreshed daily, snapshot 2026-09-19.