DeepSeek has become one of the most popular AI model families for reasoning, coding, mathematics, and general-purpose conversations. While you can access DeepSeek through online services, another option is to run a DeepSeek model directly on your Windows 11 computer. Running an AI model locally gives you more control over your data and allows you to use the model without sending every prompt to an online service.
One of the easiest ways to run DeepSeek locally on Windows 11 is with Ollama. Ollama provides a simple way to download and run large language models from the Windows desktop or command line. It also supports hardware acceleration where available, and its Windows application can provide a graphical interface for chatting with local models.
In this guide, I will show you how to install Ollama, download DeepSeek-R1, run it locally, choose an appropriate model size, and troubleshoot common problems. You don’t need to be an AI expert or know how to program to follow these steps.
Method 1: Install Ollama on Windows 11
The easiest way to run DeepSeek locally is to use Ollama as the local AI runtime. Ollama handles downloading the model, starting the model, and communicating with it.
Ollama officially supports Windows 10 and later, so it can be installed directly on Windows 11.
Step 1: Download Ollama
Open your web browser and go to the official Ollama Windows download page:
Download the Windows installer and save it to your computer.
Step 2: Install Ollama
After the download finishes:
- Open the downloaded Ollama installer.
- Follow the installation instructions.
- Allow Windows to install the application if User Account Control appears.
- Wait for the installation to complete.
- Ollama will run in the background after installation.
You don’t normally need to configure a complicated virtual machine or Linux environment. Ollama’s Windows implementation provides native support and can use compatible hardware acceleration.
Step 3: Verify the installation
Open Windows Terminal, PowerShell, or Command Prompt.
You can press:
Windows + X
and select Terminal.
Then type:
ollama --version
Press Enter.
If Ollama is installed correctly, Windows should display the installed Ollama version.
If you see an error saying that ollama is not recognized, close Terminal, open it again, and try the command once more.
Method 2: Download DeepSeek-R1
Once Ollama is installed, you can download a DeepSeek model.
Ollama currently provides several DeepSeek-R1 variants, including 1.5B, 7B, 8B, 14B, 32B, 70B and much larger versions. The different versions require significantly different amounts of storage and system resources.
For most Windows 11 computers, you should start with a smaller model rather than immediately attempting to run the largest available version.
Step 1: Open Terminal
Open Windows Terminal or PowerShell.
Then enter:
ollama run deepseek-r1
Press Enter.
Ollama will download the DeepSeek-R1 model if it isn’t already installed.
The default deepseek-r1 model is currently based on the DeepSeek-R1 0528 8B model and is listed at approximately 5.2 GB.
The download can take some time depending on your internet connection.
Step 2: Wait for the download
During the first run, Ollama downloads the model files to your computer.
Do not close the Terminal window while the model is being downloaded.
The amount of time required depends on your internet connection and disk speed.
You should also make sure you have sufficient free storage space. Although the model itself may be several gigabytes, additional disk space is useful for the operating system, Ollama, other models, and temporary files.
Step 3: Start chatting
When the model has finished downloading, Ollama will start DeepSeek.
You can then type something such as:
Explain how Windows 11 works in simple terms.
Press Enter.
DeepSeek will process the request locally on your computer.
You can ask it questions about programming, mathematics, writing, troubleshooting, general knowledge, and many other subjects.
Step 4: Exit DeepSeek
When you’re finished, press:
Ctrl + D
or close the Terminal window.
The model remains installed on your computer, so you don’t need to download it again the next time you want to use it.
Method 3: Run a Specific DeepSeek Model
One of the advantages of Ollama is that you can choose different model sizes.
For example, you can run the 1.5B version with:
ollama run deepseek-r1:1.5b
The 7B version can be started with:
ollama run deepseek-r1:7b
The 8B version can be started with:
ollama run deepseek-r1:8b
Larger versions include:
ollama run deepseek-r1:14b
ollama run deepseek-r1:32b
and:
ollama run deepseek-r1:70b
These models are progressively more demanding. Ollama lists the approximate model sizes as 1.1 GB for 1.5B, 4.7 GB for 7B, 5.2 GB for 8B, 9.0 GB for 14B, 20 GB for 32B and 43 GB for 70B.
Which model should you choose?
If you have a relatively modest Windows 11 PC, start with:
ollama run deepseek-r1:1.5b
If your computer has more RAM and a reasonably powerful processor or GPU, you can try:
ollama run deepseek-r1:7b
or:
ollama run deepseek-r1:8b
The 14B and larger versions are better suited to computers with substantially more available memory and processing power.
The important point is that a larger model isn’t automatically the best choice for every computer. A model that is too large for your hardware may run extremely slowly or cause your system to run out of memory.
Method 4: Use DeepSeek Through the Ollama Desktop App
If you don’t want to work entirely from the command line, Ollama also provides a graphical application for Windows.
Ollama’s newer Windows application allows users to download models and chat with them through a graphical interface. It also supports working with files for supported models.
After installing Ollama, open the Ollama application from the Windows Start menu.
Depending on the current version, you can use its interface to find and download models.
Search for:
DeepSeek-R1
Select the model you want and allow Ollama to download it.
After installation, you can start a conversation through the graphical interface rather than entering commands into PowerShell.
This is particularly useful if you’re introducing local AI to someone who isn’t comfortable using Terminal.
Method 5: Check Which DeepSeek Models Are Installed
You can see which models are already downloaded to your computer by running:
ollama list
Ollama will display the models available locally.
For example, you may see something similar to:
NAME SIZE
deepseek-r1:8b 5.2 GB
The exact information displayed depends on the Ollama version and the models installed.
This command is useful if you’ve downloaded several different AI models and want to determine which ones are occupying storage space.
Method 6: Remove a DeepSeek Model
If you no longer need a particular model, you can remove it.
First check the installed models:
ollama list
Then remove the model using:
ollama rm deepseek-r1:7b
Replace deepseek-r1:7b with the exact model name shown on your computer.
Removing an unused model can free several gigabytes of storage.
Be careful when using this command because the selected model and its local files will be removed. If you want to use that model again later, you will need to download it again.
Method 7: Use DeepSeek Completely Locally
One of the main reasons people install DeepSeek locally is privacy and local processing.
When you run a model through Ollama on your Windows computer, the local Ollama service is available through the computer’s localhost address. Ollama documents its local API at:
http://localhost:11434
This allows compatible applications to communicate with your locally running model.
For example, Ollama’s DeepSeek documentation provides an API example using:
http://localhost:11434/api/chat
This means local applications can interact with DeepSeek without requiring you to manually copy every prompt into a browser-based AI service.
However, “local” doesn’t mean that every part of your computer automatically becomes private. You should still be careful about the applications you install and the services you connect to your local AI environment.
Method 8: Test Whether DeepSeek Is Working Properly
After installing DeepSeek, try several simple prompts.
Start with:
Hello. Are you running locally?
Then try a reasoning task:
If a train travels 120 km in 2 hours, what is its average speed?
You can also test programming:
Write a simple Python program that calculates the factorial of a number.
If DeepSeek answers these questions successfully, your local installation is working.
For a more practical test, ask it to summarize a piece of text or help troubleshoot a Windows problem.
Troubleshooting
Ollama command is not recognized
If you see:
'ollama' is not recognized as an internal or external command
try closing Windows Terminal and opening it again.
If that doesn’t work, restart Windows and try:
ollama --version
again.
Also make sure Ollama was actually installed rather than simply downloading the installer.
DeepSeek is extremely slow
Local AI performance depends heavily on your hardware.
If DeepSeek is responding very slowly, try a smaller model.
For example, instead of:
ollama run deepseek-r1:32b
try:
ollama run deepseek-r1:7b
A smaller model can be considerably easier for a typical Windows computer to run.
Ollama supports hardware acceleration for compatible systems, including NVIDIA GPUs, while also supporting CPU-based execution using available CPU instruction sets.
Windows becomes slow while DeepSeek is running
A large language model can consume significant system resources.
If Windows becomes sluggish:
- Close unnecessary applications.
- Stop the current DeepSeek session.
- Try a smaller model.
- Check available RAM in Task Manager.
- Make sure you have sufficient free disk space.
Avoid choosing a model simply because it has a larger parameter count.
The model won’t download
Check your internet connection first.
You can also try:
ollama pull deepseek-r1
This tells Ollama to download the model without immediately starting a chat.
After the download completes, run:
ollama run deepseek-r1
Ollama’s official DeepSeek page also documents ollama pull deepseek-r1 as the command for updating an older local copy.
Not enough memory
If you receive memory-related errors or your computer becomes extremely slow, switch to a smaller model.
For example:
ollama run deepseek-r1:1.5b
is considerably smaller than:
ollama run deepseek-r1:32b
Remember that the listed model file size is not the same thing as the total amount of RAM required during operation. The operating system, Ollama, model context, and other applications also consume memory.
Frequently Asked Questions
1. Can I run DeepSeek locally without an internet connection?
Yes, after the model has been downloaded, the actual inference can be performed locally. You initially need an internet connection to install Ollama and download the model files. Once the model is stored on your computer, you can use the local model without accessing the online DeepSeek chat service.
2. How much RAM do I need to run DeepSeek on Windows 11?
There isn’t one universal RAM requirement because DeepSeek-R1 is available in multiple model sizes.
Smaller models such as the 1.5B version are much easier to run on modest hardware. Models such as 7B and 8B are more demanding, while 14B, 32B and 70B require progressively more resources.
For a normal Windows 11 PC, starting with the 7B or 8B class is generally a more practical experiment than immediately trying one of the very large models.
3. Is DeepSeek free to run locally?
The DeepSeek-R1 models available through Ollama can be downloaded and run locally without paying for each individual prompt. However, you still need suitable computer hardware, storage space, electricity, and an initial internet connection to download the software and model.
The DeepSeek-R1 model family is listed under the MIT license on Ollama’s model page, with specific licensing notes for its distilled models and their underlying model families.
4. Can I use DeepSeek for coding?
Yes. DeepSeek-R1 is designed as a reasoning model and can be used for programming, mathematics, logic, writing, and many other tasks. Ollama also provides an API that allows compatible software applications to communicate with a locally running model.
Conclusion
Running DeepSeek locally on Windows 11 is much easier than it may initially appear. With Ollama, you don’t need to manually configure a complicated AI environment or download a large collection of dependencies. Install Ollama, download a suitable DeepSeek-R1 model, and start chatting with it from Windows Terminal or the Ollama application.
The simplest approach is:
ollama run deepseek-r1
If your computer has limited hardware, start with a smaller model such as:
ollama run deepseek-r1:1.5b
or:
ollama run deepseek-r1:7b
If your computer has considerably more memory and processing power, you can experiment with larger versions such as 14B, 32B, or 70B.
The biggest advantage of running DeepSeek locally is having an AI model available directly on your own computer. Ollama also exposes a local API, making it possible to connect compatible applications to your locally running model.
For beginners, the best approach is to start small, verify that everything works, and then move to a larger model if your hardware can handle it. This lets you experience local AI on Windows 11 without unnecessarily consuming your computer’s memory or storage.
