DeepSeek-R1: Requirements and Deployment
DeepSeek-R1 is a state-of-the-art reasoning model that has set new benchmarks in complex problem-solving, particularly in mathematics, science, and coding. Its performance is comparable to OpenAI’s O1 model and is available under the MIT license, promoting open-source collaboration and commercial use.
Model Variants and Hardware Requirements
DeepSeek-R1 comes in various versions, including full-scale models and distilled variants optimized for different hardware capabilities.
Full-Scale Models
- DeepSeek-R1 and DeepSeek-R1-Zero
- Parameters: 71 billion
- VRAM Requirement: ~1,342 GB
- Recommended Setup: Multi-GPU configuration, such as 16 NVIDIA A100 GPUs with 80GB each
Distilled Models
These versions are optimized to retain significant reasoning capabilities while reducing hardware demands.
| Model | Parameters (B) | VRAM Requirement (GB) | Recommended GPU |
|---|---|---|---|
| DeepSeek-R1-Distill-Qwen-1.5B | 1.5 | ~0.7 | NVIDIA RTX 3060 12GB or higher |
| DeepSeek-R1-Distill-Qwen-7B | 7 | ~3.3 | NVIDIA RTX 3070 8GB or higher |
| DeepSeek-R1-Distill-Llama-8B | 8 | ~3.7 | NVIDIA RTX 3070 8GB or higher |
| DeepSeek-R1-Distill-Qwen-14B | 14 | ~6.5 | NVIDIA RTX 3080 10GB or higher |
| DeepSeek-R1-Distill-Qwen-32B | 32 | ~14.9 | NVIDIA RTX 4090 24GB |
| DeepSeek-R1-Distill-Llama-70B | 70 | ~32.7 | NVIDIA RTX 4090 24GB (x2) |
Running DeepSeek-R1 Locally
For users without access to high-end multi-GPU setups, the distilled models offer a practical alternative. These models can be run on consumer-grade hardware with varying VRAM capacities.
Using Ollama
Ollama is a tool that facilitates running open-source AI models locally.
Installation
- Download and install Ollama from the official website.
Model Deployment
Run the 8B distilled model using the following command:
ollama run deepseek-r1:8b
For other model sizes, replace 8b with the desired model parameter size (e.g., 1.5b, 14b).
API Interaction
Start the Ollama server:
ollama serve
Send requests using curl:
curl -X POST http://localhost:11434/api/generate -d '{
"model": "deepseek-r1",
"prompt": "Your question or prompt here"
}'
Replace "Your question or prompt here" with your actual input prompt.
Conclusion
DeepSeek-R1 offers a range of models to accommodate various hardware configurations. While the full-scale models require substantial computational resources, the distilled versions provide accessible alternatives for users with limited hardware capabilities. Tools like Ollama simplify the process of running these models locally, enabling a broader audience to leverage advanced reasoning capabilities.