OpenCompass Subjective - Reasoning - Chinese (CompassBench 2405): leaderboard
Metric: Score (%). Source: rank.opencompass.org.cn. 39 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4o (2024-05-13) | 57.8 |
| 2 | GPT-4 Turbo | 57 |
| 3 | GLM-4 | 44.1 |
| 4 | Claude 3 Opus | 43.8 |
| 5 | Moonshot v1 8K | 42.1 |
| 6 | Yi Large | 40.1 |
| 7 | DeepSeek V2 Chat | 39.6 |
| 8 | Yi 1.5 34B Chat | 37 |
| 9 | Qwen 1.5 110B Chat | 36.3 |
| 10 | Llama 3 70B Instruct | 33.2 |
| 11 | Mistral Large | 32.2 |
| 12 | internlm2-chat-20B | 30.9 |
| 13 | Qwen 1.5 72B Chat | 30.7 |
| 14 | Qwen 1.5 32B Chat | 29.1 |
| 15 | Yi-1.5-9B Chat | 27.6 |
Interactive version: theaggregate.ai/benchmark?slug=opencompass-subjective-reasoning-chinese-compassbench-2405 · How It Works · Data refreshed daily, snapshot 2026-09-19.