Since its release, the DeepSeek-R1 model has attracted the attention of many users. In order to allow deepin users to better experience this cutting-edge technology, UOS AI has now integrated the DeepSeek-R1 edge-side model!

Users can download the deepin - adapted UOS AI DeepSeek - R1 1.5B model to use DeepSeek - R1 within UOS AI.

Note: This method supports deepin 23 and deepin 25 Preview systems.

Step 1: Update UOS AI and Embedding Plugin

• For deepin 23 Users:
  You can update UOS AI to the latest version from the application store.

• For deepin 25 Preview Users:
  Please use the following command to upgrade UOS AI to the latest version.

sudo apt update
sudo apt install uos-ai deepin-modelhub uos-ai-rag

Note: If this is your first time downloading and using UOS AI on deepin, follow the on-screen prompts to "Accept the User Agreement - Get a Free Trial Account" to access the latest version of UOS AI directly.

After updating, open the Settings option in UOS AI, go to Model Configuration - Local Models, and select Embedding Plugin - Install/Update. You will be automatically redirected to the application store interface, where you can click Install/Update to complete the process.

Step 2: Install the DeepSeek-R1 GGUF Model

The backend framework of UOS AI is based on llama.cpp, and you can directly find the GGUF version of DeepSeek-R1 on the Model Community. Here, we will use the most common and convenient 1.5B model as an example. The model size is moderate, and it does not require high hardware specifications; it can run smoothly with just a CPU. (If you are using CUDA for inference, you can wait for the deepin system to update the CUDA version of the inference backend framework.)

The download links for the DeepSeek-R1 1.5B model, which has been adapted for UOS AI, are as follows:

• Baidu Drive: https://pan.baidu.com/s/1-786vFU4dunpdproiOHdbQ?pwd=zxxw
• 123 Drive: https://www.123865.com/s/XXsfjv-Jjcph
• 123 Drive(Alternative link): https://www.123684.com/s/XXsfjv-Jjcph)
• Quark Drive: https://pan.quark.cn/s/edbf721fefaf

After downloading the model file, open the file manager and go directly to the following directory:

/home/username/.local/share/deepin-modelhub/models   #Replace "username" with your own computer username.

Note: If the deepin-modelhub folder and the models folder within it do not exist, you can manually create these folders.

Enter the models folder, extract the downloaded model folder, and directly copy it to the path of the newly created folder. Make sure to check that there are no extra names in the path. The complete path of the model is as follows:

/home/username/.local/share/deepin-modelhub/models/DeepSeek-R1-1.5B/gguf    #Replace "username" with your own computer username.

After completing the above steps, open the UOS AI interface again and you can manually switch to use the DeepSeek-R1-1.5B model.