SeaEval - Cultural Reasoning - CN-Eval (Zero-Shot): leaderboard
Metric: Accuracy (%). Source: huggingface.co. 25 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 2.5 72B Instruct | 87.62 |
| 2 | Qwen 2.5 32B Instruct | 84.76 |
| 3 | Qwen 2 7B Instruct | 82.86 |
| 4 | Qwen 2.5 14B Instruct | 82.86 |
| 5 | Qwen 2 72B Instruct | 82.86 |
| 6 | SeaLLMs-v3-7B-Chat | 81.9 |
| 7 | Qwen 2.5 7B Instruct | 80 |
| 8 | Qwen 2.5 3B Instruct | 71.43 |
| 9 | Sailor2-8B-Chat | 71.43 |
| 10 | gemma2-9B-cpt-sea-lionv3-instruct | 59.05 |
| 11 | Gemma 2 9B (IT) | 58.1 |
| 12 | Qwen 2.5 1.5B Instruct | 55.24 |
| 13 | Llama 3.1 70B Cpt Sea Lionv3 Instruct | 55.24 |
| 14 | Llama 3.1 70B Instruct | 54.29 |
| 15 | Llama 3 70B Instruct | 53.33 |
Interactive version: theaggregate.ai/benchmark?slug=seaeval-cultural-reasoning-cn-eval-zero-shot · How It Works · Data refreshed daily, snapshot 2026-09-05.