OpenCompass Agent - T-Eval - Chinese (CompassBench 2405): leaderboard
OpenCompass CompassBench 2405 edition (file dated 2024-06-12): general tool utilization (T-Eval) sub-task of the agent category, Chinese questions. CompassBench is OpenCompass's periodic evaluation of chat models on a private question set; each edition is a separate board. Score 0-100, higher is better.
Metric: Score (%). Source: rank.opencompass.org.cn. Saturation forecast: Estimated already saturated. 39 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Nanbeige2-16B-Chat | 90.4 | #971 |
| 2 | Mixtral 8x22B Instruct | 85.4 | #738 |
| 3 | GPT-4 Turbo | 84.9 | #417 |
| 4 | Mistral Large | 84.9 | #713 |
| 5 | Yi Large | 84.6 | #566 |
| 6 | Claude 3 Opus | 83.9 | #565 |
| 7 | Command-R+ | 83.1 | #869 |
| 8 | GPT-4o (2024-05-13) | 81.9 | #337 |
| 9 | Yi 1.5 34B Chat | 79.4 | #798 |
| 10 | Qwen 1.5 32B Chat | 75.9 | #820 |
| 11 | DeepSeek V2 Chat | 75.6 | #572 |
| 12 | Mistral 7B Instruct (v0.3) | 74.3 | #1255 |
| 13 | GLM-4 | 74.2 | #647 |
| 14 | GPT-3.5 Turbo | 72.5 | #858 |
| 15 | DeepSeek-V2-Lite-Chat | 71.2 | #1189 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=opencompass-agent-t-eval-chinese-compassbench-2405 · How It Works · Data refreshed daily, snapshot 2026-11-05.