OpenCompass LLM - Agent - English (CompassBench 2402): leaderboard
Metric: Score (%). Source: rank.opencompass.org.cn. 35 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4 Turbo (Preview) | 84.4 |
| 2 | GPT-3.5 Turbo | 77.3 |
| 3 | GLM-4 | 77.3 |
| 4 | Qwen Max | 69.7 |
| 5 | Qwen 1.5 14B Chat | 66 |
| 6 | Qwen-72B-Chat | 63.8 |
| 7 | internlm2-chat-20B | 57.7 |
| 8 | Qwen-14B-Chat | 57.5 |
| 9 | internlm2-chat-7B | 54.9 |
| 10 | Yi 34B (Chat) | 53.9 |
| 11 | Mixtral 8x7B Instruct (v0.1) | 53.7 |
| 12 | Qwen-7B Chat | 53.4 |
| 13 | Qwen 1.5 7B Chat | 52.8 |
| 14 | Qwen 1.5 4B Chat | 52.7 |
| 15 | WizardLM-13B-V1.2 | 47.4 |
Interactive version: theaggregate.ai/benchmark?slug=opencompass-llm-agent-english-compassbench-2402 · How It Works · Data refreshed daily, snapshot 2026-09-19.