Build Your Own Private "ChatGPT" Server: The 2026 Guide to DeepSeek & Ollama

20 Min Read | No GPU Required

Stop paying $20/month for AI that spies on you. In 2026, you don't need a $10,000 server to run smart AI. Thanks to "Quantization," you can run powerful models like DeepSeek-V3 or Llama 4 on a cheap standard VPS.

This guide will teach you how to build a 100% private, uncensored AI chatbot that you can access from your phone or laptop anywhere in the world.

1. The Hardware (It's Cheaper Than You Think)

We are using CPU Inference. It's slightly slower than a GPU, but infinitely cheaper.

OS: Ubuntu 24.04 or 26.04 (LTS)
RAM: 8GB minimum (for 7B/8B models). 16GB recommended for larger models.
Storage: 50GB NVMe (Models are large files).

Don't have a server? Check our VPS offers here.

2. Install The Foundation: Docker

We will use Docker containers. This keeps your AI software clean and isolated from the rest of your server. Run these commands one by one:

curl -fsSL https://get.docker.com -o get-docker.sh
sudo sh get-docker.sh

3. The Brain: Installing Ollama

Ollama is the backend magic. It translates the AI model files into something your computer can understand. We will run it as a Docker container.

docker run -d -v ollama:/root/.ollama -p 11434:11434 --name ollama ollama/ollama

Now, let's download a "Brain". We will use the DeepSeek-R1 Distilled (8 Billion parameters) model. It's fast and incredibly smart for coding.

docker exec -it ollama ollama run deepseek-r1:8b

Note: It will download about 5GB of data. Once you see a prompt, type "Hello" to test it, then press `Ctrl+D` to exit the chat.

4. The Face: Open WebUI

Ollama is just a command line. You want a ChatGPT-like interface. Enter Open WebUI. It looks exactly like ChatGPT, supports dark mode, and saves your chat history.

docker run -d -p 3000:8080 --add-host=host.docker.internal:host-gateway -v open-webui:/app/backend/data --name open-webui --restart always ghcr.io/open-webui/open-webui:main

Wait 30 seconds, then open your browser and go to: http://YOUR_SERVER_IP:3000

Create an admin account (the first account created becomes the Admin). You will see your DeepSeek model ready to select in the top left corner!

5. Access From Anywhere (Cloudflare Tunnel)

Right now, your AI is only accessible via IP address, which is insecure (HTTP). Let's give it a secure https://ai.yourdomain.com address without opening risky ports.

A. Install Cloudflared

curl -L --output cloudflared.deb https://github.com/cloudflare/cloudflared/releases/latest/download/cloudflared-linux-amd64.deb
sudo dpkg -i cloudflared.deb

B. Create the Tunnel

Login: cloudflared tunnel login
Create Tunnel: cloudflared tunnel create private-ai
Route Traffic: Map your domain to the tunnel. Replace UUID with the ID given in the previous step.

cloudflared tunnel route dns private-ai ai.yourdomain.com

C. Run It

Tell the tunnel to forward traffic to your Open WebUI port (3000).

cloudflared tunnel run --url http://localhost:3000 private-ai

Result

You can now open https://ai.yourdomain.com on your phone. You have your own private DeepSeek app that costs $0/month (after server costs) and nobody reads your data.