## 模型文件规格

| 文件名 | 量化类型 | 文件大小 | 描述 |
| -------- | ---------- | --------- | ----------- |
| [deepseek-r1-14b-cot-math-reasoning-full-Q2_K.gguf](https://huggingface.co/tensorblock/deepseek-r1-14b-cot-math-reasoning-full-GGUF/blob/main/deepseek-r1-14b-cot-math-reasoning-full-Q2_K.gguf) | Q2_K | 5.770 GB | 最小尺寸，显著质量损失 - 不推荐大多数用途 |
| [deepseek-r1-14b-cot-math-reasoning-full-Q3_K_S.gguf](https://huggingface.co/tensorblock/deepseek-r1-14b-cot-math-reasoning-full-GGUF/blob/main/deepseek-r1-14b-cot-math-reasoning-full-Q3_K_S.gguf) | Q3_K_S | 6.660 GB | 极小尺寸，高质量损失 |
| [deepseek-r1-14b-cot-math-reasoning-full-Q3_K_M.gguf](https://huggingface.co/tensorblock/deepseek-r1-14b-cot-math-reasoning-full-GGUF/blob/main/deepseek-r1-14b-cot-math-reasoning-full-Q3_K_M.gguf) | Q3_K_M | 7.339 GB | 极小尺寸，高质量损失 |
| [deepseek-r1-14b-cot-math-reasoning-full-Q3_K_L.gguf](https://huggingface.co/tensorblock/deepseek-r1-14b-cot-math-reasoning-full-GGUF/blob/main/deepseek-r1-14b-cot-math-reasoning-full-Q3_K_L.gguf) | Q3_K_L | 7.925 GB | 小尺寸，较大质量损失 |
| [deepseek-r1-14b-cot-math-reasoning-full-Q4_0.gguf](https://huggingface.co/tensorblock/deepseek-r1-14b-cot-math-reasoning-full-GGUF/blob/main/deepseek-r1-14b-cot-math-reasoning-full-Q4_0.gguf) | Q4_0 | 8.518 GB | 传统方案；小尺寸，极高质量损失 - 建议优先使用Q3_K_M |
| [deepseek-r1-14b-cot-math-reasoning-full-Q4_K_S.gguf](https://huggingface.co/tensorblock/deepseek-r1-14b-cot-math-reasoning-full-GGUF/blob/main/deepseek-r1-14b-cot-math-reasoning-full-Q4_K_S.gguf) | Q4_K_S | 8.573 GB | 小尺寸，较大质量损失 |
| [deepseek-r1-14b-cot-math-reasoning-full-Q4_K_M.gguf](https://huggingface.co/tensorblock/deepseek-r1-14b-cot-math-reasoning-full-GGUF/blob/main/deepseek-r1-14b-cot-math-reasoning-full-Q4_K_M.gguf) | Q4_K_M | 8.988 GB | 中等尺寸，平衡质量 - 推荐 |
| [deepseek-r1-14b-cot-math-reasoning-full-Q5_0.gguf](https://huggingface.co/tensorblock/deepseek-r1-14b-cot-math-reasoning-full-GGUF/blob/main/deepseek-r1-14b-cot-math-reasoning-full-Q5_0.gguf) | Q5_0 | 10.267 GB | 传统方案；中等尺寸，平衡质量 - 建议优先使用Q4_K_M |
| [deepseek-r1-14b-cot-math-reasoning-full-Q5_K_S.gguf](https://huggingface.co/tensorblock/deepseek-r1-14b-cot-math-reasoning-full-GGUF/blob/main/deepseek-r1-14b-cot-math-reasoning-full-Q5_K_S.gguf) | Q5_K_S | 10.267 GB | 大尺寸，低质量损失 - 推荐 |
| [deepseek-r1-14b-cot-math-reasoning-full-Q5_K_M.gguf](https://huggingface.co/tensorblock/deepseek-r1-14b-cot-math-reasoning-full-GGUF/blob/main/deepseek-r1-14b-cot-math-reasoning-full-Q5_K_M.gguf) | Q5_K_M | 10.509 GB | 大尺寸，极低质量损失 - 推荐 |
| [deepseek-r1-14b-cot-math-reasoning-full-Q6_K.gguf](https://huggingface.co/tensorblock/deepseek-r1-14b-cot-math-reasoning-full-GGUF/blob/main/deepseek-r1-14b-cot-math-reasoning-full-Q6_K.gguf) | Q6_K | 12.125 GB | 超大尺寸，极低质量损失 |
| [deepseek-r1-14b-cot-math-reasoning-full-Q8_0.gguf](https://huggingface.co/tensorblock/deepseek-r1-14b-cot-math-reasoning-full-GGUF/blob/main/deepseek-r1-14b-cot-math-reasoning-full-Q8_0.gguf) | Q8_0 | 15.702 GB | 超大尺寸，极低质量损失 - 不推荐 |

## 下载指南

### 命令行方式

首先安装Huggingface客户端：

```shell
pip install -U "huggingface_hub[cli]"

然后下载单个模型文件到本地目录：

huggingface-cli download tensorblock/deepseek-r1-14b-cot-math-reasoning-full-GGUF --include "deepseek-r1-14b-cot-math-reasoning-full-Q2_K.gguf" --local-dir 本地目录路径

如需批量下载符合特定模式的文件（例如*Q4_K*gguf），可使用：

huggingface-cli download tensorblock/deepseek-r1-14b-cot-math-reasoning-full-GGUF --local-dir 本地目录路径 --local-dir-use-symlinks False --include='*Q4_K*gguf'