DeepSeek is deployed locally, with visualization, no delay, and smooth use
Easily build a local DeepSeek and enjoy a latency-free, highly private intelligent conversation experience.
Core content:
1. DeepSeek API service latency issues and the necessity of local deployment
2. Local deployment hardware requirements for different versions of DeepSeek-R1 models
3. Detailed steps to build a local DeepSeek intelligent assistant using Ollama and Chatbox
The first step is to select the model
The following are the hardware resources required for local deployment of different versions of DeepSeek-R1 distillation models. We will download the corresponding models based on our own hardware resources later:
Step 2: Download and run the model
2.1 Download Ollama (URL: https://ollama.com/download)
Ollama is used to run AI models locally and is very easy to use.
Taking Windows system as an example, after installing OllamaSetup.exe, double-click the software icon to run the Ollama service.
2.2 Use Ollama to download and run the model. Here we choose the smallest 1.5b model (1.5 billion parameters)
ollama run deepseek-r1:1.5b
This is the command to run the model. Since there is no model locally, Ollama will download the model from the warehouse before running it, which is equivalent to:
ollama pull deepseek-r1:1.5b ollama run deepseek-r1:1.5b
At this point, the model service has been started. The default local service address is http://localhost:11434. With this local API service address, we can configure it to other software to create a local DeepSeek smart assistant, such as a code assistant. See DeepSeek Code Assistant.
You can also use it directly from the command line:
Step 3: Configure the dialogue interface
If you want a chat interface like the large models commonly seen on the market, you can use Chatbox
3.1 Install Chatbox (URL: https://chatboxai.app/zh)
