What DeepSeek is and how to get it running

DeepSeek is an open-source AI model you can run locally on your own computer, rather than using it through a web browser or paying a service. Installation means downloading the model files and the software that runs them, then setting them up so you can chat with the AI from your machine. The process takes 30 minutes to an hour depending on your internet speed and which installation method you choose.

Running DeepSeek locally means your conversations stay on your device — they don't go to a company's server. It also means you can use it offline once it's installed. The trade-off is that you need enough disk space (roughly 8 to 40 gigabytes depending on the model size) and your computer does the thinking work instead of a remote server, so responses are slower than cloud-based AI tools.

Key Takeaways

  • DeepSeek runs through software like Ollama or LM Studio, which you download first, then use to pull the model files from the internet.
  • You need at least 8 GB of free disk space for the smallest DeepSeek model, and 16 GB or more for larger versions that give better answers.
  • Windows, Mac, and Linux all support DeepSeek through the same tools, though setup steps differ slightly between operating systems.
  • After installation, you interact with DeepSeek through a text interface on your computer or through a web browser window that runs locally.

Installing Ollama, the easiest path for most people

Ollama is the simplest way to get DeepSeek running. It handles downloading the model and provides a text interface you can use immediately. Go to ollama.ai, download the installer for your operating system (Windows, Mac, or Linux), and run it like you would any other program.

Once Ollama is installed, open your terminal or command prompt. On Windows, search for "Command Prompt" in the Start menu. On Mac, open Spotlight (Command + Space) and type "Terminal". On Linux, open your terminal application. Type this command and press Enter:

ollama run deepseek-r1:7b

This downloads the 7-billion-parameter version of DeepSeek, which is the smallest and fastest on most home computers. The download is about 4 GB and may take 10 to 30 minutes depending on your internet speed. Once it finishes, you can start typing questions and the AI will respond in your terminal window.

If you want a more capable version that gives longer, more detailed answers, you can instead run ollama run deepseek-r1:32b, but this requires about 20 GB of disk space and a computer with at least 16 GB of RAM to run smoothly.

Using LM Studio for a graphical interface

If you prefer clicking buttons instead of typing commands, LM Studio provides a desktop application with a chat window. Download it from lmstudio.ai, install it, and open the application.

On the left side, you'll see a search box. Type "deepseek" and the available models appear. Click the download button next to "deepseek-r1:7b" (or whichever size you want). The model downloads to your computer — this takes the same 10 to 30 minutes as Ollama, depending on your internet.

Once the download finishes, click the chat icon at the bottom left. Select the DeepSeek model from the dropdown menu at the top, type your question in the text box, and press Enter. LM Studio shows the AI's response in the chat window above.

Checking your computer's disk space and RAM

Before you start downloading, check whether your computer has enough room. The 7-billion-parameter model needs about 4 to 5 GB of free disk space after installation. The 32-billion-parameter model needs 18 to 20 GB. The 70-billion-parameter model needs 40 GB or more and runs only on computers with 32 GB of RAM or higher.

On Windows, right-click your C: drive, select Properties, and look at the "Free space" number. On Mac, click the Apple menu, select "About This Mac", click "Storage", and see how much space is available. On Linux, open a terminal and type df -h to see available space.

For RAM, on Windows open Task Manager (Ctrl + Shift + Esc), click the Performance tab, and look at Memory. On Mac, open Activity Monitor (Command + Space, type "Activity Monitor"), click the Memory tab, and look at the number next to "Swap Used" — if it's high, your computer is struggling. On Linux, type free -h in a terminal.

If your computer has less than 8 GB of RAM, the 7-billion-parameter model will run very slowly. If you have 8 to 16 GB, the 7-billion model works fine. If you have 16 GB or more, you can try the 32-billion model.

Troubleshooting common installation problems

If the download starts but stops partway through, your internet connection dropped. Close the terminal or application and run the command again — Ollama and LM Studio resume where they left off instead of starting over.

If you get an error saying "command not found" after installing Ollama on Mac or Linux, the installer didn't add Ollama to your system path. Restart your computer and try again. If that doesn't work, open a new terminal window and type /usr/local/bin/ollama run deepseek-r1:7b on Mac, or ollama run deepseek-r1:7b on Linux after checking that Ollama is in your PATH.

If LM Studio opens but won't download the model, check your internet connection by opening a web browser and visiting a website. If the web works but LM Studio doesn't, try closing and reopening LM Studio. If it still fails, uninstall LM Studio, restart your computer, and reinstall it.

If the AI responds very slowly (taking a minute or more per sentence), your computer doesn't have enough RAM for the model size you chose. Try the smaller 7-billion model instead, or close other programs to free up memory.

Running DeepSeek without a terminal after installation

After you've installed DeepSeek through Ollama, you can use it through a web interface instead of typing in a terminal. Open a new terminal window and type:

ollama serve

Leave this window open. Then open your web browser and go to http://localhost:3000. You'll see a chat interface where you can type questions. This window stays open as long as the terminal window is running.

Alternatively, download Open WebUI from openwebui.com for a more polished chat interface. Install it, run it, and it connects to your local DeepSeek installation automatically. You get a Gmail-like chat experience with the ability to save conversations and organize them into folders.

Understanding model sizes and what they mean for you

DeepSeek comes in different sizes: 1.5 billion, 7 billion, 32 billion, and 70 billion parameters. The number refers to how many mathematical weights the model learned during training — more parameters generally means better answers, but also slower responses and higher disk and RAM requirements.

The 7-billion model is the sweet spot for most home computers. It answers straightforward questions quickly, handles coding tasks reasonably well, and runs on machines with 8 GB of RAM. The 32-billion model gives noticeably better answers for complex questions and creative writing, but takes 2 to 3 times longer to respond and needs 16 GB of RAM minimum. The 70-billion model is the most capable but impractical for most home setups.

You can install multiple sizes on the same computer if you have the disk space. Run ollama run deepseek-r1:7b one day and ollama run deepseek-r1:32b another — they don't interfere with each other.

Frequently Asked Questions

Can I use DeepSeek offline after I install it?

Yes. Once the model is downloaded, you can run DeepSeek without an internet connection. Ollama and LM Studio work entirely locally. You only need internet during the initial download of the model files.

Is DeepSeek free to use after installation?

Yes. DeepSeek is open-source, and Ollama and LM Studio are free software. There are no subscription fees or usage limits once it's on your computer. You're only paying for the electricity to run it.

How do I uninstall DeepSeek if I change my mind?

On Windows and Mac, uninstall Ollama or LM Studio like any other program (Control Panel on Windows, Applications folder on Mac). The model files are stored in a folder on your computer — on Windows it's usually in AppData, on Mac it's in .ollama. Uninstalling the software doesn't automatically delete these files, so you may need to manually delete the folder to free up disk space.

Can I run DeepSeek on my phone or tablet?

Not easily. DeepSeek is designed for computers with significant processing power. Some Android phones can run smaller models through apps like Termux, but it's slow and requires technical knowledge. For phones and tablets, using DeepSeek through a web browser (accessing it on someone else's computer or a cloud service) is more practical.

What's the difference between deepseek-r1 and other DeepSeek versions?

DeepSeek-R1 is the reasoning-focused version that shows its thinking process before answering. Other versions like DeepSeek-Chat are optimized for conversation. For most people, deepseek-r1 is the better choice because it explains its reasoning, which helps you understand whether the answer is reliable.