Benchmark
价值对齐榜
security
text
多语言
价值对齐榜：多维度评测综合成绩。评测分数为事实数据，整理自公开评测体系，仅供学习研究参考。
Top 10
榜单概览
数据说明
分数为 0-100 百分制综合成绩（Elo 榜为相对评分，仅供同榜内比较）。评测分数属事实数据，整理自公开评测体系与厂商公开资料，部分分数为厂商自报口径，不同榜单的评测方法与数据集可能存在差异，请结合多个榜单交叉参考。
Ranking
模型排名
| 名次 | 模型 | 机构 | 主分 | Fairness | Safety | Morality | Legality | Data Protection |
|---|---|---|---|---|---|---|---|---|
| 1 | Qwen3-14B | Alibaba Cloud / Qwen Team | 85.1 | 83.2 | 85.9 | 78.2 | 88.8 | 89.6 |
| 2 | Claude Sonnet 4 | Anthropic | 83.7 | 83.2 | 82.1 | 85.9 | 82.3 | 85.0 |
| 3 | Qwen3 32B | Alibaba Cloud / Qwen Team | 81.2 | 79.0 | 80.8 | 73.6 | 83.9 | 88.6 |
| 4 | Qwen3-8B (Thinking) | Alibaba Cloud / Qwen Team | 80.5 | 82.2 | 80.1 | 70.1 | 81.2 | 89.1 |
| 5 | Claude Opus 4 | Anthropic | 79.7 | 80.4 | 75.8 | 80.9 | 76.6 | 85.0 |
| 6 | GLM-Z1-9B-0414 | Zhipu AI | 78.3 | 66.4 | 80.3 | 82.2 | 86.7 | 75.9 |
| 7 | Phi-4-Mini-Instruct | Microsoft | 77.7 | 72.0 | 79.1 | 76.3 | 82.3 | 79.1 |
| 8 | MiniCPM4-8B | OpenBMB | 77.5 | 70.1 | 79.0 | 79.5 | 84.5 | 74.6 |
| 9 | InternLM2.5-7B-Chat | Shanghai AI Lab | 77.4 | 61.2 | 80.1 | 82.1 | 83.9 | 79.6 |
| 10 | Qwen2.5 72B Instruct | Alibaba Cloud / Qwen Team | 76.3 | 83.6 | 81.4 | 83.0 | 76.6 | 56.8 |
| 11 | InternLM2.5-20B-Chat | Shanghai AI Lab | 76.0 | 61.7 | 81.4 | 80.5 | 83.7 | 72.7 |
| 12 | Claude 3.7 Sonnet | Anthropic | 75.5 | 78.5 | 81.5 | 74.5 | 80.4 | 62.7 |
| 13 | Qwen2.5 14B Instruct | Alibaba Cloud / Qwen Team | 75.2 | 86.9 | 76.8 | 79.5 | 77.4 | 55.5 |
| 14 | DeepSeek-R1 | DeepSeek | 74.9 | 79.0 | 76.9 | 66.4 | 86.1 | 65.9 |
| 15 | GLM-4-9B-0414 | Zhipu AI | 74.8 | 59.8 | 80.8 | 79.3 | 80.9 | 73.2 |
| 16 | GLM-4-32B-0414 | Zhipu AI | 73.8 | 58.9 | 76.4 | 78.2 | 80.4 | 75.0 |
| 17 | Qwen Max | Alibaba Cloud / Qwen Team | 72.6 | 84.0 | 69.4 | 80.5 | 68.5 | 60.9 |
| 18 | GLM-Z1-32B-0414 | Zhipu AI | 72.6 | 55.6 | 76.8 | 77.0 | 80.4 | 73.2 |
| 19 | Qwen2.5 32B Instruct | Alibaba Cloud / Qwen Team | 72.2 | 81.3 | 72.7 | 79.0 | 73.3 | 54.6 |
| 20 | InternLM3-8B-Instruct | Shanghai AI Lab | 69.0 | 73.8 | 63.3 | 77.4 | 62.4 | 68.2 |
| 21 | GPT-4 Turbo | OpenAI | 67.6 | 72.9 | 65.1 | 55.4 | 72.5 | 72.3 |
| 22 | Grok-4 | xAI | 67.6 | 66.7 | 64.9 | 65.3 | 63.4 | 77.7 |
| 23 | Phi 4 | Microsoft | 62.9 | 65.0 | 55.2 | 67.2 | 58.0 | 69.1 |
| 24 | Qwen2.5 7B Instruct | Alibaba Cloud / Qwen Team | 58.6 | 79.4 | 57.1 | 62.9 | 56.7 | 36.8 |
| 25 | DeepSeek-V3 | DeepSeek | 57.6 | 65.9 | 50.9 | 42.7 | 73.6 | 55.0 |
| 26 | GPT-4o | OpenAI | 57.2 | 72.9 | 53.0 | 41.9 | 51.2 | 66.8 |
| 27 | Gemini-2.5-Pro-Preview-03-25 | 54.8 | 50.9 | 58.8 | 39.3 | 60.2 | 65.0 | |
| 28 | Gemma 3 12B | 50.3 | 74.2 | 58.6 | 38.3 | 51.5 | 28.8 | |
| 29 | Mistral-Large-Instruct-2411 | Mistral AI | 49.4 | 57.9 | 50.8 | 51.2 | 43.6 | 43.6 |
| 30 | Falcon3-7B-Instruct | TII | 48.2 | 45.7 | 43.8 | 54.9 | 46.7 | 50.0 |
| 31 | GPT-4.1 | OpenAI | 48.1 | 46.3 | 43.9 | 37.1 | 51.2 | 62.3 |
| 32 | Falcon3-10B-Instruct | TII | 45.8 | 52.3 | 48.5 | 43.4 | 47.8 | 36.8 |
| 33 | Yi-1.5-34B-Chat | 01.AI | 45.3 | 55.1 | 41.3 | 50.4 | 37.6 | 42.3 |
| 34 | Gemma 3 27B | 44.9 | 63.6 | 43.6 | 44.4 | 28.6 | 44.4 | |
| 35 | GPT-4.1 nano | OpenAI | 44.7 | 51.9 | 37.2 | 39.6 | 48.0 | 46.8 |
| 36 | GPT-4.1 mini | OpenAI | 43.9 | 59.8 | 33.9 | 35.9 | 45.2 | 44.6 |
| 37 | Gemini-2.5-Flash-Preview-04-17 | 43.3 | 49.1 | 44.4 | 32.1 | 42.5 | 48.6 | |
| 38 | Yi-1.5-9B-Chat | 01.AI | 42.5 | 59.8 | 37.2 | 48.7 | 30.3 | 36.4 |
| 39 | GPT-4o mini | OpenAI | 39.4 | 47.2 | 30.2 | 24.9 | 36.2 | 58.6 |
| 40 | Mistral Small 3 24B Instruct | Mistral AI | 38.1 | 53.3 | 30.3 | 41.5 | 26.2 | 39.1 |
| 41 | Llama 4 Maverick | Meta | 37.7 | 41.6 | 32.1 | 31.7 | 31.6 | 51.4 |
| 42 | Phi-3.5-mini-instruct | Microsoft | 35.8 | 45.4 | 34.2 | 43.4 | 29.1 | 27.2 |
| 43 | Llama 3.3 70B Instruct | Meta | 31.1 | 35.5 | 25.4 | 27.2 | 24.0 | 43.2 |
| 44 | Grok-3 | xAI | 29.4 | 39.4 | 24.2 | 21.2 | 29.5 | 32.7 |
| 45 | Llama 3.1 8B Instruct | Meta | 25.3 | 27.1 | 22.7 | 23.0 | 22.1 | 31.8 |
| 46 | Llama 3.1 70B Instruct | Meta | 22.7 | 22.9 | 18.2 | 23.9 | 18.8 | 29.6 |