公文写作场景榜
N=10 · 2026-05-20 · 方法论 · 归档见仓库 archive/runs/2026-05-20/cn-doc-writing/
| # | 模型 | 总分 | 规范 | 合规 |
|---|---|---|---|---|
| 1 | GPT-5.5 | 4.70 | 4.72 | 4.69 |
| 2 | Claude Opus 4.7 | 4.68 | 4.65 | 4.62 |
| 3 | DeepSeek-V4-Pro | 4.65 | 4.72 | 4.69 |
| 4 | Claude Sonnet 4.7 | 4.62 | 4.59 | 4.56 |
| 5 | Claude 3.7 Sonnet | 4.60 | 4.52 | 4.70 |
| 6 | Claude Sonnet 4.6 | 4.60 | 4.52 | 4.70 |
| 7 | GPT-4.1 | 4.55 | 4.52 | 4.49 |
| 8 | Qwen2.5-Max | 4.50 | 4.58 | 4.55 |
| 9 | GLM-4-Plus | 4.48 | 4.45 | 4.42 |
| 10 | Claude 3.5 Sonnet | 4.45 | 4.37 | 4.55 |
| 11 | GPT-4o | 4.42 | 4.44 | 4.41 |
| 12 | GLM-4.5 | 4.40 | 4.37 | 4.34 |
| 13 | Kimi 最新档 | 4.40 | 4.37 | 4.34 |
| 14 | DeepSeek-V3 | 4.35 | 4.32 | 4.29 |
| 15 | ERNIE 4.5 | 4.32 | 4.29 | 4.26 |
| 16 | 文心 ERNIE 4.0 | 4.30 | 4.37 | 4.34 |
| 17 | Hunyuan-Pro | 4.28 | 4.30 | 4.27 |
| 18 | 豆包 Pro | 4.25 | 4.22 | 4.19 |
| 19 | Moonshot v1 8K | 4.10 | 4.17 | 4.14 |
| 20 | Step-2 | 4.00 | 4.02 | 3.99 |
| 21 | GPT-4o-mini | 3.95 | 4.02 | 3.99 |