A massive event just rocked the AI community—DeepSeek has gone viral worldwide! Thanks to its free and open-source nature and output quality rivaling OpenAI's O1 model, it has gained a huge following. Just take a look at GitHub—its starred count has skyrocketed to over 60K, up from 16K the last time we checked. That’s insane!
However, we’ve noticed that DeepSeek’s online version sometimes experiences server overload, making it hard to access. So today, we’re bringing you a step-by-step tutorial on how to run DeepSeek R1 locally and offline. Let’s dive right in!
Step 1: Install Ollama
First, we need to install a runtime environment that allows local execution. We recommend Ollama, which supports running all versions of the DeepSeek R1 model for free.
Head over to the official website, click the download button, and select the version compatible with your computer. For this example, we'll use a Windows PC. The file size is around 700MB. Once downloaded, double-click to install. After installation, you'll see it in your system tray—that means you're good to go!!
Step 2: Copy the DeepSeek R1 Model Code
Now, let’s get the DeepSeek R1 model:
- Go back to Ollama’s official website and click the "Models" tab at the top left.
- The first option you see should be DeepSeek R1—that’s right, it’s already integrated!
- Click on the model, and you’ll find different versions based on your computer’s VRAM capacity.
- If you’re unsure which one to choose, here’s a general recommendation:Lower VRAM? Pick a smaller model.
- Slow inference speed? Downgrade to a lower version.
- For this tutorial, we’ll use the 7B version, which is the most commonly used.
- Click the copy button next to it—this will copy the command we need to download the model automatically.
Step 3: Open Windows PowerShell
Now, let’s install the model:
- Find Windows PowerShell in your system’s search bar and open it.(Or press WIN + R, type cmd, and hit Enter to open the command prompt.)
- Paste the command you copied earlier and hit Enter.
- Ollama will now automatically download and install DeepSeek R1 for you.
- Once the installation is complete, you can start asking questions directly!
- Try typing a question and pressing Enter—you should see "Think" appear, meaning DeepSeek R1 is processing your query.
However, we understand that this command-line interface may not be user-friendly for beginners. So, if you prefer a clean and intuitive UI—something like ChatGPT—let’s move on to the next step.
Step 4: Run DeepSeek R1 Using the Page Assist Extension
To enhance your experience, let’s set up a browser-based interface:
- Open the Chrome Web Store and search for "Page Assist."
- Click "Add to Chrome."—The first result is the one you want
- Once installed, find the Page Assist icon in the top-left corner. Open it, and you'll see it's already linked to the model and Ollama we just set up.
- Select the model you installed earlier. Before using it, we need to tweak some basic settings.
- Click the settings icon in the top-right corner, enable "RAG Settings," and select your model under "Embedding Model."
- Save your changes, and you're ready to go! Now, try asking a sample question:
Boom! The answer appears instantly.
Additionally, you can also enable the internet search feature in the bottom-left corner to ask about the latest news. Pretty straightforward, right? You can even switch between other models if you have them installed!
Final Thoughts for DeepSeek
This is how you can run DeepSeek R1 locally and offline using Ollama and the Page Assist extension. If you have any questions, leave a comment on AIPURE YouTube! And don’t forget to like, subscribe, and turn on notifications for our channel so you don’t miss our future tutorials.
