OpenCompass Agent - T-Eval - Chinese (CompassBench 2407): leaderboard
OpenCompass CompassBench 2407 edition (file dated 2024-08-05): 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. 36 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Nanbeige2-16B-Chat | 90.3 | #971 |
| 2 | Yi Large | 85.6 | #566 |
| 3 | Mixtral 8x22B Instruct | 85.5 | #738 |
| 4 | Llama 3.1 70B Instruct | 85.1 | #567 |
| 5 | Gemma 2 27B (IT) | 85 | #652 |
| 6 | GPT-4o Mini (2024-07-18) | 84.7 | #568 |
| 7 | Qwen 2 72B Instruct | 84.1 | #579 |
| 8 | Claude 3.5 Sonnet (20240620) | 83.4 | #308 |
| 9 | Mistral Large 2 (Jul) | 82.1 | #490 |
| 10 | GPT-4o (2024-05-13) | 81.8 | #337 |
| 11 | DeepSeek V2 Chat (0628) | 81.2 | #409 |
| 12 | Mistral Large | 79.6 | #713 |
| 13 | internlm2-chat-20B | 79.5 | #988 |
| 14 | Llama 3.1 8B Instruct | 78.3 | #1017 |
| 15 | InternLM2.5-Chat-7B | 78.1 | #1009 |
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-2407 · How It Works · Data refreshed daily, snapshot 2026-11-05.