The new Chinese AI model DeepSeek R1 has created ripples in the AI industry. The model, which claims to outperform OpenAI's o1 model, can be accessed for free from their website. However, there are also a lot of privacy concerns involved. If you're also harboring such concerns, you can instead run the model locally on your device.

There are a few different models available for DeepSeek R1 including the distilled models with 1.5B, 7B, 14B, 32B, and 70B parameters and the full flagship model with 671B parameters. To run the higher-end models, i.e., the 14B and above models, you'll need a machine with more memory and a faster CPU and GPU, and perhaps, even a dedicated NPU.

However, even to run the lowest models, i.e., the 1.5B and 7B models, you need at least 8 GB of RAM.

Run DeepSeek R1 locally on PC with LM Studio

LM Studio is one of the simplest ways to run DeepSeek R1 locally on your Windows, Mac, or Linux PC.

LM Studio can run on the following systems:

  • Macs: Apple Silicon Chips (M1, M2, M3, M4) with macOS 13.4 or later
  • Windows: Both x64 and ARM (Snapdragon X Elite) are supported
  • Linux: Ubuntu 20.04 or newer, and x64 is required

While the official recommendation is 16 GB RAM, you can run the smaller models with 8 GB RAM but expect some trouble. If you have more than 192 GB of RAM, you can even run the full 671B flagship DeepSeek R1 model which is not distilled.

  1. Now, go to lmstuido.ai and click on the 'Download' button corresponding to your system.
  1. Follow the instructions as per your system to install the app.
  2. Open LM Studio once it is installed and click on 'Skip onboarding'.
  1. Next, click on the 'Search' icon from the navigation menu on the left.
Note: If you can't see the Search icon, either switch to the 'Power User' user type from the bottom. Alternatively, go to the 'App Settings' icon from the bottom-right corner and then navigate to 'Model Search' in the Mission Control window that opens if you don't want to change your user type.
  1. Then, search for the model you want to download. You can install the DeepSeek R1 Distill (Qwen 7B), for instance, if you want to use the model with 7B parameters or the DeepSeek R1 Distill Qwen 1.5B model for 1.5B parameters.

    All DeepSeek R1 models are available for download as long as you have enough space on your system as listed on the download page. Click on the 'Download' button to begin.
  1. Once the model is downloaded, click on the 'Load Model' button and the chat window will open.
  1. Simply enter your message in the window and chat with the model locally. You'll be able to monitor the resource usage by the model from the bottom-right corner of the status bar.
  1. To change any settings for the model, click on the 'Setting' icon on the left of the model name at the top.

LM Studio is one of the easiest ways to use DeepSeek R1 locally on your PC where you don't even need to use the terminal. There are many other options available as well, like using Ollama, but that involves the use of the terminal.

Run DeepSeek R1 locally on phone with PocketPal AI

There are many apps available that can let you run LLM models locally on your mobile phone. My recommendation? PocketPal AI. It is a simple app to use LLM models locally on your iOS or Android device.

  1. Install the app from the App Store or the Play Store.
  2. Then, open the app and tap on 'Go to Models'.
  3. Tap on the '+' icon at the bottom right corner of the screen.
  1. Next, tap on 'Add from Hugging Face'.
  2. Type 'DeepSeek R1' in the search bar to find the model you want to install and tap on it.
  1. Then, tap on the download icon on one of the available GGU files.
  2. Once the download is complete, go back to the main screen and tap on the 'Load' icon on the downloaded model.
  1. The model will be loaded and you can start chatting with it.
  2. You can find your chat history or change the model settings from the menu on the left.

There you go. No matter your device, using the apps listed above, you can install and run DeepSeek R1 models locally with ease.