← All models
Hugging FaceModelScope0 clicks 1.78B684.49B685.36B
DeepSeek-R1
以推理链闻名的 R1，MIT 许可可商用
deepseek-ai684.49BMoE中文ChatReasoning
LicencemitCommercial use OK
- Parameters
- 684.49B
- Context
- 160K
- Architecture
- MoE
- Released
- Jan 20, 2025
Run it
Copy and paste; the tool pulls the weights itself.
llama-cli -hf deepseek-ai/DeepSeek-R1 vllm serve deepseek-ai/DeepSeek-R1
Downloadable versions
Sizes and VRAM are estimates from the parameter count; the link opens the official repository.
| Format | Quant | Size | Min VRAM | Download |
|---|---|---|---|---|
| gguf | Q3_K_M | 306 GB | 379 GB | Hugging Face |
| awq | W4A16 | 351 GB | 434 GB | Hugging Face |
| gguf | Q4_K_M | 370 GB | 457.5 GB | Hugging Face |
| gguf | Q5_K_M | 453 GB | 560 GB | Hugging Face |
| gguf | Q6_K | 523 GB | 646.5 GB | Hugging Face |
| gguf | Q8_0 | 676 GB | 835.5 GB | Hugging Face |
| safetensors | BF16 | 1.25 TB | 1575.5 GB | Hugging Face |
Similar models
DeepSeek-R1-Distill-Qwen-1.5B
deepseek-ai
R1 蒸馏到 1.5B，小显卡也能跑推理模型
中文ChatReasoning128K
Commercial use OKfrom 983 MB
DeepSeek-V3
deepseek-ai
DeepSeek 第三代基础模型，685B MoE
MoE中文Chat160K
Restricted licencefrom 306 GB
DeepSeek-V3.2
deepseek-ai
DeepSeek 第三代 MoE 旗舰，MIT 许可
MoE中文ChatReasoning160K
Commercial use OKfrom 306 GB