What is DeepSeek and how to Run it Locally on your Computer

Recently, DeepSeek, an AI model by a Chinese hedge fund company has taken the world by storm. It has crashed around 1 trillion dollars worth of tech stocks. Including $600 Billion from Nvidia alone. All of this being said, what is Deep Seek and why is it so revolutionary?

Essentially Deepseek is a startup led by Liang Wenfeg that recently released their new R1 model. It is funded and owed by the Chinese hedge fund, High-Flyer. Their new R1 model has outperformed OpenAI's o1 model in many metrics. This means that DeepSeek's free model is better than ChatGPT's paid model. These metrics were based on how accurately the AI models could solve math coding and English questions. Here is a table of R1's performance compared to other models.

Credit: Prompthub.us

As seen above Deepseek can outperform OpenAI's models in many different benchmarks. The best part is that DeepSeek took a fraction of the cost to train compared to OpenAI's models. DeepSeek's model took around 6 million dollars whilst OpenAI's model cost around $100 million. In addition, DeepSeek requires far less powerful hardware to run as well as being far more environmentally friendly. This is why Nvidia stocks have fallen so much if powerful hardware is not needed to run AI then Nvidia GPUs won't be as useful anymore. That being said running the full version of the DeepSeek R1 model locally still requires multiple multiple RTX 4090s, however, OpenAI's models require massive servers. The best part is that it is that DeepSeek is open-source which means all of its code is public and free to use. In this article, we will be learning how to run a distilled version of DeepSeek that does not require multiple 4090s. Without further to do let's get started.

Pros and Cons of Running DeepSeek Locally

Pros: Running R1 locally means that your data won't be collected by DeepSeek nor will it be sold. It also means that you can run the model on your computer without the Internet.

Cons: Running AI models locally also means using resources from your computer. This means that if you don't have a very good PC it could make it more difficult to run these models.

Recommended Hardware Requirments

• 16 GB of RAM

• 4.68 GB Disk Space

• GTX 960

Keep in mind that you don't exactly need 16 GB of RAM to run the model I only had 12. But of course the more RAM the better. The same goes for GPUs the more powerful the better, it's ok if you have a weaker GPU than the 960. Keep in mind that if you have less powerful hardware, the model may take more time to respond to your prompts compared to having more powerful hardware.

Step #1

Go to https://lmstudio.ai/ and download the LM-Studio app on your device, depending on whether you have a Mac, Linux, or Windows. Also, make sure to choose whether you are an x86 architecture or Arm.

Step #2

Click on the .exe or .dmg file that you downloaded and go through the steps required to install the app. This step may take a good few minutes.

Step #3

Now that you have installed the app, open it up and click the Discover button on the left edge (right image). Once you have clicked that button you will see a column full of AI models that you can download. Choose the DeepSeek R1 Distill (Qwen 7B) (bottom) and click the green download icon at the bottom right of the screen. Then your model will start downloading. It may take a good few minutes depending on your internet speed and hardware.

Step #4

Now that you have downloaded the model you can run it locally by simply clicking the Purple 'Use in Chat Button' at the bottom right of the screen. Once you press the button, you can chat with the model just like how you would chat with ChatGPT. Keep in mind, depending on your hardware, getting responses from the model could take quite a bit of time. With a mid-range laptop from 4 years ago, It took the model around 4 minutes on average to answer my prompts. However, if you have better hardware it could take less time. Feel free to download other AI models and play around.

Conclusion DeepSeek R1 is a game-changer in AI, offering top-tier performance at a fraction of the cost. Its open-source nature, efficiency, and lower hardware demands make AI more accessible while disrupting the tech industry. Running it locally enhances privacy but requires decent hardware. As AI advances, DeepSeek R1 marks a shift toward more efficient and widely available AI solutions.