Deploy, Invoke, and Try DeepSeek on d.run¶

d.run currently offers a free one-week trial for DeepSeek models.
Visit https://console.d.run/ to try out the following DeepSeek models with one click:

• DeepSeek-R1: The full-powered version with 236 billion parameters. It uses a dense Transformer architecture and activates all modules during every inference.
  It demonstrates strong capabilities in complex reasoning tasks, especially excelling in math reasoning, code generation, and logical deduction.
  In these high-end application domains, R1 is considered on par with top-tier models like OpenAI's GPT-4, offering much stronger reasoning power than typical general-purpose models.

• DeepSeek-V3: Based on a Mixture of Experts (MoE) architecture with 671 billion parameters, activating 37 billion per inference.
  It performs excellently on general natural language processing tasks and stands out in terms of response speed.

• DeepSeek-R1-Distill-Qwen-32B: A distilled model.

• DeepSeek-R1-Distill-Qwen-14B: A distilled model.

However, the free trial usually only lasts one week. It is recommended to deploy your own exclusive large model.
The following uses DeepSeek-R1-Distill-Qwen-14B as an example.

Deploying the Model¶

1. In the Model Plaza, find DeepSeek-R1-Distill-Qwen-14B and click the Deploy button on the card.

   

2. Enter the basic information and resource details, select a billing method, then click Confirm.

   

3. When the system prompts "Deployment Successful" and the status changes from Deploying to Running,
   it means the DeepSeek-R1-Distill-Qwen-14B model has been successfully deployed.

   

Chat with the Deployed Model in drun¶

In the image above, click Experience in the operations column to start chatting with the DeepSeek-R1-Distill-Qwen-14B distilled model.

Calling the d.run Model from Third-party Apps¶

You can also call the d.run API from third-party intelligent apps such as VSCode, Bob Translate, Lobe Chat, Cherry Studio, etc.
This allows you to use the just-deployed DeepSeek-R1-Distill-Qwen-14B model.

Creating an API Key¶

To call d.run's model services in a third-party app, you’ll need an API Key.

1. In the d.run LLM service platform, go to Analytics & Management -> API Key Management, then click Create on the right.

   

2. Enter an easily recognizable name, then click Confirm.

   

3. When prompted that the API Key was created successfully, securely save the generated key and click Close.

   

4. Return to the API Key list. The newly generated key appears at the top by default.

   

Calling d.run Model Service¶

Using the third-party app Bob Translate as an example:

1. After installing it via the App Store, launch Bob and select Preferences from the dropdown menu.

   

2. Click Services -> ➕

   

3. Select DeepSeek.

   

4. Fill in the following parameters, enable only d.run’s model service, then click Save.

   Config Item	Example Value
   Service Name	d.run.deepseek
   API Key	Enter the newly generated Key
   Custom API Base URL	https://sh-02.d.run
   Custom API Path	Leave empty or /v1/chat/completions
   Model	Custom model
   Custom Model	DeepSeek-R1-Distill-Qwen-14B

   

   Note

   You can also run the following command in the terminal to check if the model is callable:

   curl 'https://sh-02.d.run/v1/chat/completions' \
    -H "Content-Type: application/json" \
    -H "Authorization: Bearer <replace with your API Key>" \
    -d '{
       "model": "u-3d7a8e49da2a/test14b",
       "messages": [{"role": "user", "content": "Say this is a test!"}],
       "temperature": 0.7
    }'
   

5. Now try letting Bob Translate translate a sentence.

   

🎉 Congratulations! The third-party app Bob Translate has successfully called the DeepSeek-R1-Distill-Qwen-14B model deployed in d.run via the API.
Similarly, you can integrate any DeepSeek model deployed in d.run with any third-party intelligent application.