DeepSeek R1 Distill Qwen 7B
开源来自 DeepSeek · 发布于 2025-01-20
32.7
平均分
N/A
输入价格
N/A
输出价格
N/A
上下文窗口
text-generation
类型
Tested on 11 benchmarks with 32.7% average. Top scores: JSQuAD (74.2%), JCommonsenseQA (59.8%), JNLI (54.6%).
基准测试分数
| 基准测试 | 类别 | 分数 | Bar |
|---|---|---|---|
| JSQuAD | language | 74.2 | |
| JCommonsenseQA | language | 59.8 | |
| JNLI | language | 54.6 | |
| JMMLU | language | 42.3 | |
| IFEval | language | 40.4 | |
| LLM-JP — Overall | language | 39.3 | |
| MATH Level 5 | math | 19.6 | |
| MMLU-PRO | knowledge | 14.7 | |
| BBH (HuggingFace) | general | 7.9 | |
| GPQA | knowledge | 3.9 | |
| MUSR | reasoning | 3.5 |