How to Run DeepSeek Locally on Your Windows PC and Mac

DeepSeek AI has taken the technology industry by storm in the US. This Chinese AI startup skyrocketed to the number one position in the Apple App Store’s best free app category, overtaking OpenAI’s ChatGPT. The DeepSeek AI also rattled the US tech stock market by wiping off almost $400 billion from NVIDIA’s market cap. The company then also launched the Janus-Pro image generation model that surpassed OpenAI’s Dall-E model. If that wasn’t impressive, it was reported that DeepSeek was able to offer all of these top-notch offerings for a fraction of the cost compared to its competitors, such as OpenAI.

While DeepSeek AI is available to be accessed for free from their official website, there are privacy concerns attached to it since it originates from China. If you are confused about how DeepSeek will use your data and send it to China, then you might want to run this AI chatbot locally on your computer. In this guide, we will share with you the details of how to run DeepSeek AI locally on Windows and Mac. Let’s jump right into it.

What is DeepSeek R1?

DeepSeek AI is another AI chatbot that comes from China. It was founded by Liang Wenfeng in December 2023, and the DeepSeek R1 model, their AI chatbot, was introduced in January 2025. It works similarly to other chatbots, such as Microsoft’s Copilot or OpenAI’s ChatGPT. While you might use it for the same tasks, such as generating content, asking questions, writing or summarizing content, or assisting with coding tasks, the difference between DeepSeek and other AI models lies in the former’s efficiency.

Reportedly, only $6 million was invested to train DeepSeek, compared to the over $100 million invested by OpenAI for ChatGPT. DeepSeek AI can produce better results compared to its competitors; however, it is trained to avoid politically sensitive questions. When we repeatedly tried to ask DeepSeek AI questions about “Arunachal Pradesh,” an unalienable part of India, the AI chatbot refused to provide any answers. So, that’s there.

Otherwise, the chatbot produces effectively better results when asked to code or generate content. As soon as DeepSeek rose to popularity, it was under scrutiny for handling the data of users and how it was able to obtain high-performance NVIDIA chips to power the R1 model, as the US government restricts it for exports.

Run DeepSeek locally on Windows and Mac computers

The best part about the below methods is that the steps are the same for both Windows and Mac and even Linux computers. The software that we will be using is also available for free for both Windows and Mac.

Use LM Studio

1. Visit the official website of LM Studio and download the free software on your Windows or Mac PC.
2. After downloading and installing LM Studio, launch the program.
3. Click on the search icon on the left pane.
4. On the left pane, under Model Search, locate the DeepSeek R1 Distill (Qwen 7B) model. It weighs about 4.68GB in size. Hit the Download button at the bottom.
5. Do note that you would need at least 5GB of free storage space and 8GB RAM on your PC to download this model.
6. After completing the DeepSeek AI model download, head over to the Chat window and load the model.
7. Select the model and click on the Load Model button. In case you come across an issue, you should dial back the GPU offload to 0 and then try again.
8. That’s it! You can now chat and interact with DeepSeek AI locally on your PC without worrying about your data getting uploaded to Chinese servers.

Use Ollama

1. Download and install Ollama software from the official website on your Windows, Mac, or Linux computer.
2. Launch Command Prompt on your Windows PC or Terminal on your Mac computer, and run the below command:
   • ollama run deepseek-r1:1.15b
3. The above code is for an efficient version of DeepSeek R1 and is based on Qwen for low-end computers. It only makes use of 1.1GB of memory. However, if you have a high-end PC with powerful specs, then you can copy code for other DeepSeek AI models such as 7b, 8b, 14b, 32b, and 70b from below:
   • 7b (uses 4.7GB of memory)—ollama run deepseek-r1:7b
   • 8b (uses 4.9GB of memory)—ollama run deepseek-r1:8b
   • 14b (uses 9GB of memory)—ollama run deepseek-r1:14b
   • 32b (uses 20GB of memory)—ollama run deepseek-r1:32b
   • 70b (uses 43Gb of memory)—ollama run deepseek-r1:70b
4. This will load the DeepSeek AI model inside the Command Prompt. You are all set to ask questions or use it any way you like.
5. If you wish to stop your conversation with the AI model, you need to type in the Ctrl+D keyboard shortcut.

Use DeepSeek AI locally and keep your data private

Running DeepSeek AI locally using the above two methods ensures that your data isn’t sent to external servers and is available locally. Also, you will notice faster responses and also the option to integrate DeepSeek AI into any workflow. However, do note that if you own a powerful computer, then you should use the main DeepSeek R1 model. Those using an old or underpowered PC can choose to use the DeepSeek 1.5b model. Another thing that we would like to point out is that upon our testing, we found multiple responses, particularly about history, wrong and fabricated, which is one of the main reasons you shouldn’t rely much on AI.