How to Install and Run Ollama on a VPS (2026 Guide)

Yes, you can run Ollama on a VPS, and I have done it myself, multiple times, across different VPS configurations. 

I have personally tested every command in this guide on a live Ubuntu 24.04 VPS, pulled models ranging from 1B to 70B and verified the API, Nginx proxy and security setup. 

I have added notes at every step to make sure you do not miss a thing.

You already have a VPS and you want to run AI models on it and you do not want to send your data to a third-party API.

I will tell you every step. I will tell you how to connect to your VPS, install Ollama, pull an AI model, run it, and expose it safely over the internet.

By the time you finish this guide, your VPS will be serving a live AI model API you can call from anywhere. And in doing all that:

NO GPU is required at all. A CPU-only VPS with 4 GB of RAM is enough to get started.

What Is Ollama & Why Run Ollama on VPS?

Ollama is an open-source runtime that lets you download and run large language models (LLMs) on your own hardware using a single command. 

I have been working with VPS infrastructure for years and Ollama is the cleanest self-hosted LLM solution I’ve come across. It handles model downloads, runtime management and serves an OpenAI-compatible REST API as well.

Popular models you can run with Ollama include:

• Llama 3 
• Mistral 
• Gemma
• DeepSeek 
• Qwen
• Phi

All of them run locally and your prompts and responses never leave your server.

But if you CAN run Ollama locally on your computer (PLEASE DO THAT ON YOUR RISK), then why to run Ollama on a VPS? Why is a VPS better than your local machine?

See! The thing is that running Ollama locally is fine for quick experiments. 

The moment you want to keep the model running 24/7 or you want to access it from your phone or you want to share it with a teammate without exposing your home IP, your local machine stops making sense and a VPS becomes the only choice.

You can even integrate Ollama into a web app or automation workflow and also keep sensitive data off third-party APIs when you run Ollama on VPS.

So here is what you gain the moment you move Ollama to a VPS. Every one of these I have verified in testing environments.

• Zero API costs: No per-token billing, at all
• Full data privacy: Prompts and responses stay on your server
• Always-on availability: 24/7 uptime, no sleep mode
• Remote access: Call your model from any device, anywhere
• Scalable: upgrade RAM and CPU in just one click as soon as the need grows

VPS Requirement for Ollama: What You Need Before You Start

Before you type a single command, make sure your VPS meets the minimum requirements. I have run Ollama on both low-end and high-end VPS instances. 

Believe me! The biggest problem is RAM. The entire model must fit in memory or speed drops really very badly.

Minimum Hardware Requirements

Use this table to check your VPS against the minimum and recommended specs before you begin.

How Much RAM Does Each Model Need?

The table below shows approximate RAM requirements for quantized Q4 models. Full-precision models need two to three times more RAM than these figures. 

Start with a 3B or 7B model.

And, what about the operating system? The OS? Ubuntu 22.04 LTS or Ubuntu 24.04 LTS is the recommended OS. The Ollama install script handles everything automatically on both versions. 

All commands in this guide assume you are running Ubuntu. With your VPS ready and SSH credentials in hand, let’s go step by step.

Step Zero: Get Your VPS from YouStable

Before you run a single command in this guide, you need a VPS. I recommend YouStable and here is exactly why it fits Ollama perfectly.

Every plan includes Ubuntu 24.04 LTS support (required for this guide), full root SSH access, advanced DDoS protection, free migration, and 99.95% uptime, exactly what you need for a 24/7 Ollama deployment.

Now, Let’s discuss how to buy a YouStable VPS: Step by Step

• Go to the official YouStable VPS page >> youstable.com/vps-hosting

• Choose your plan based on the RAM table above >> For beginners, start with vStart (4 GB) or vPro (6 GB) >> Click “Buy Now” on your chosen plan.

Select your billing cycle (Choose Annual as it saves more).

• Choose your data center location >> Select USA for global access or India for Indian traffic >> Select Ubuntu 24.04 LTS as your operating system.

• Complete checkout >> Your VPS login credentials and IP address will be emailed to you.

Use those credentials to SSH into your VPS and follow this guide from Step 1.

Step 1: Connect to Your VPS via SSH

Everything you do with Ollama is done through the terminal. You connect to your VPS using SSH. This is how every VPS is managed, and it is the first skill you need to have before anything else.

What Is SSH and Why You Need It

SSH gives you a secure, encrypted connection from your local computer to your remote VPS. Once connected, you see a command prompt that controls the remote server, exactly as if you were sitting in front of it.

How to Connect: Linux and macOS

Open your terminal and run the following command. Replace YOUR_VPS_IP with the IP address from your VPS control panel.

ssh root@YOUR_VPS_IP

If your provider gave you a non-root user, which is common on Ubuntu, use that username instead:

ssh ubuntu@YOUR_VPS_IP

How to Connect: Windows

On Windows 10 or later, open PowerShell or Command Prompt and run the same ssh command above. You can also use PuTTY: enter your VPS IP in the Host Name field, select port 22, and click Open.

First Login: Accept the Host Key

The first time you connect, you will see a message asking you to confirm the host fingerprint. Type yes and press Enter. This is normal and expected.

The authenticity of host ‘1.2.3.4’ can’t be established.

Are you sure you want to continue connecting (yes/no)?

Step 2: Update Your VPS System

Before installing anything, update the system package list. I always do this first on every fresh VPS. It ensures you are installing the latest, most secure versions of all dependencies and avoids version conflicts later.

apt update && apt upgrade -y

This command does two things: apt update fetches the latest list of available packages from Ubuntu’s repositories, and apt upgrade -y installs all pending updates automatically. The -y flag confirms all prompts without asking you.

This step typically takes one to two minutes. Wait for it to finish before proceeding.

Step 3: Install Ollama on Your VPS

Ollama provides an official installer script that handles everything. 

It detects your OS, downloads the correct binary, creates an ollama system user, and sets up a systemd service so Ollama starts automatically on boot. I have run this script on dozens of VPS instances and it has never failed on a clean Ubuntu install.

Run the Install Script

Run this single command. That is all it takes to get Ollama fully installed and running as a background service.

curl -fsSL https://ollama.com/install.sh | sh

The script downloads the binary from the official Ollama release page and installs it to /usr/local/bin/ollama. It also configures Ollama to run as a background service managed by systemd.

Verify the Installation

Check that the Ollama service is running after the install script completes. This is the first thing I verify on every setup.

systemctl status ollama

You should see active (running) in the output. This means Ollama is running in the background and listening on port 11434 on localhost. You can also verify the installed version:

ollama --version

If both commands return output without errors, Ollama is successfully installed on your VPS.

Step 4: Pull Your First AI Model

After installation, Ollama has no models downloaded. You need to pull one before you can use it. Models are pulled from the official Ollama library at ollama.com/library. 

I recommend starting with a small model to verify everything works before pulling larger ones.

Recommended Models for Beginners on VPS

For a VPS with 4 to 8 GB of RAM and no GPU, these are the models I personally recommend starting with. Each one has been tested and runs well on entry-level hardware.

How to Pull a Model

Run the pull command for your chosen model. Ollama downloads the model file directly to your VPS. The download size ranges from 1.5 GB to 5 GB depending on the model.

ollama pull llama3.2:3b

Wait for the progress bar to complete before running the model. Do not interrupt the download.

Manage Your Downloaded Models

These two commands let you see what is downloaded and remove models to free disk space when needed.

ollama list          (# list all downloaded models)ollama rm llama3.2:3b   (# remove a specific model)

Step 5: Run the Model and Test It

Once the model is downloaded, you can start using it immediately. 

Ollama gives you two ways to interact with it: a terminal chat session and a REST API. I test both every time I set up a new instance to confirm everything is wired up correctly.

Terminal Chat Mode

To open an interactive chat session in your terminal, run the command below. You will see a prompt where you can type and receive responses directly in the terminal.

ollama run llama3.2:3b

You will see a prompt like >>> . Type your message and press Enter. To exit the chat, type /bye or press Ctrl+D.

Test the REST API

Ollama automatically starts an API server on http://localhost:11434. Test it without leaving the terminal using curl. This is the command I use every time to confirm the API is working.

curl http://localhost:11434/api/generate -d '{"model": "llama3.2:3b", "prompt": "What is a VPS?", "stream": false}'

You will get a JSON response with the model’s reply. If you see a JSON object with a response field, everything is working correctly.

curl http://localhost:11434

This should return Ollama is running. That confirms the API endpoint is active and reachable from within the VPS.

Step 6: Keep Ollama Running as a Background Service

Ollama is already set up as a systemd service during installation. This means it starts automatically when your VPS reboots. Here are the key commands to manage it. I keep these in my notes for every deployment.

Service Management Commands

Use these commands to start, stop, restart, and monitor the Ollama service on your VPS at any time.

sudo systemctl status ollama # check if runningsudo systemctl start ollama # start the servicesudo systemctl stop ollama  # stop the servicesudo systemctl restart ollama   # restart after config changessudo systemctl enable ollama # start automatically on boot

Check Logs

If Ollama is not behaving as expected, check the service logs. The -f flag streams new log entries in real time so you can watch what is happening live.

journalctl -u ollama -f

Press Ctrl+C to stop watching the log stream.

Step 7: Enable Remote Access to the Ollama API

By default, Ollama only accepts connections from localhost. This means you can only call the API from within the VPS itself. To access it from your laptop, browser, or another server, you need to expose the API safely.

I always use the Nginx reverse proxy approach in production.

Option A: Bind Ollama to All Interfaces (Simple but Less Secure)

This makes Ollama directly accessible on port 11434 from any IP. Only use this on a private network or for testing purposes when you fully understand the risks.

sudo systemctl edit ollama

Add these lines in the editor:

[Service]

Environment=”OLLAMA_HOST=0.0.0.0:11434″

Save the file, then reload and restart:

sudo systemctl daemon-reloadsudo systemctl restart ollama

Now open port 11434 in your VPS firewall:

sudo ufw allow 11434/tcp

You can now call the API from outside at http://YOUR_VPS_IP:11434

Option B: Nginx Reverse Proxy with HTTPS (Recommended)

This is the proper production approach. Keep Ollama on localhost and put Nginx in front of it to handle HTTPS, authentication, and rate limiting. This is what I use on every production deployment.

Install Nginx and Certbot

Run this command to install Nginx and the Certbot plugin for automatic SSL certificate management.

sudo apt install nginx certbot python3-certbot-nginx -y

Create the Nginx Site Configuration. Create a new config file. Replace api.yourdomain.com with your actual domain name throughout this configuration.

sudo nano /etc/nginx/sites-available/ollama

Paste the following configuration into the file:

server {    listen 80;    server_name api.yourdomain.com;     location / {        proxy_pass http://127.0.0.1:11434;        proxy_set_header Host $host;        proxy_set_header X-Real-IP $remote_addr;        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;        proxy_set_header X-Forwarded-Proto $scheme;        proxy_http_version 1.1;        proxy_set_header Upgrade $http_upgrade;        proxy_set_header Connection "upgrade";        proxy_buffering off;        proxy_cache off;        proxy_read_timeout 600s;        proxy_send_timeout 600s;    }}

Enable the site and test the configuration:

sudo ln -s /etc/nginx/sites-available/ollama /etc/nginx/sites-enabled/sudo nginx -tsudo systemctl reload nginx

 Add SSL with Certbot

Get a free SSL certificate from Let’s Encrypt. Make sure your domain’s DNS A record points to your VPS IP address before running this command.

sudo certbot --nginx -d api.yourdomain.com

Follow the prompts. Certbot will automatically update your Nginx config to redirect HTTP to HTTPS and install the certificate. Your Ollama API will now be available at https://api.yourdomain.com.

Step 8: Secure Your Ollama Deployment

A default Ollama installation exposed to the public internet is a serious security risk. I have seen this firsthand: according to Shodan, thousands of Ollama servers worldwide have port 11434 open with no authentication. 

Anyone who finds your endpoint can use your server resources, pull models, or flood the API. Securing your setup takes five minutes.

Configure UFW Firewall. Set up UFW to block port 11434 from the outside while allowing SSH, HTTP, and HTTPS traffic through. 

Always run these commands in this exact order.

sudo ufw allow 22/tcpsudo ufw allow 80/tcpsudo ufw allow 443/tcpsudo ufw deny 11434sudo ufw enablesudo ufw status

Add Basic Authentication to Nginx

Install htpasswd and create a password file. Then add authentication to your Nginx config so every request to your Ollama API requires a username and password.

sudo apt install apache2-utils -ysudo htpasswd -c /etc/nginx/.ollama_passwd your_username

Add these two lines inside the location block in your Nginx config:

auth_basic "Restricted";auth_basic_user_file /etc/nginx/.ollama_passwd;

Reload Nginx to apply the changes:

sudo systemctl reload nginx

Security Checklist

Go through this checklist before you consider your Ollama deployment production-ready. I verify every item on this list before handing off any deployment.

• UFW enabled with port 11434 blocked from outside
• SSH key authentication enabled (disable password login)
• HTTPS via Let’s Encrypt certificate in place
• Basic auth or API key added to Nginx
• Ollama bound to localhost only (OLLAMA_HOST not set to 0.0.0.0)
• Regular system updates applied (apt update && apt upgrade)

Install Open WebUI: A Browser Interface for Your Models

If you prefer a visual chat interface instead of the terminal or raw API calls, Open WebUI gives you a ChatGPT-style browser interface that connects directly to your Ollama server. 

It runs in Docker and takes about five minutes to set up. I use it on every internal deployment.

Install Docker >> Run these three commands to install Docker and configure it to start automatically with the system.

curl -fsSL https://get.docker.com | shsudo systemctl start dockersudo systemctl enable docker

Run Open WebUI

Run the following Docker command to pull and start Open WebUI. The –restart always flag ensures it comes back up automatically after a reboot.

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

Open WebUI will be accessible at http://YOUR_VPS_IP:3000. 

On your first visit, create an admin account. It will automatically connect to your Ollama instance and show all your downloaded models.

Troubleshooting Common Issues

Here are the most common problems people run into when setting up Ollama on a VPS, and exactly how to fix each one. I have encountered and resolved every issue listed here during my own deployments.

“Connection refused” When Calling the API

This usually means Ollama is not running. Check the service status and start it if it shows inactive or failed.

systemctl status ollamasudo systemctl start ollama

If it keeps failing, check the last 50 lines of logs for the actual error:

journalctl -u ollama --no-pager -n 50

Model Is Very Slow

On a CPU-only VPS, LLM generation is slower than GPU-accelerated setups. Typical speeds are 3 to 10 tokens per second for a 3B to 7B model. These steps help maximize your available performance.

• Use a smaller model: try llama3.2:1b or qwen2.5:3b instead of 7B models
• Check if the model is fully in RAM: run “free -h” and compare available RAM to model size
• Close other memory-heavy processes to free up RAM

“Error: model not found”

You need to pull the model before you can run it. Use ollama list to see all models that are currently downloaded.

ollama pull MODEL_NAMEollama list

Nginx Returns 502 Bad Gateway

This means Nginx is running but cannot reach Ollama. Work through this checklist in order until the issue is resolved.

• Ollama is running: systemctl status ollama
• OLLAMA_HOST is not set to 0.0.0.0 when using Nginx proxy (keep it on localhost)
• proxy_pass in Nginx points to http://127.0.0.1:11434
• Nginx config is valid: sudo nginx -t

VPS Runs Out of RAM While Loading Model

The model size exceeds your available RAM. Switch to a smaller model, upgrade your VPS RAM tier, or add swap space as a temporary buffer.

sudo fallocate -l 4G /swapfilesudo chmod 600 /swapfilesudo mkswap /swapfilesudo swapon /swapfile

Use Cases: What You Can Build With Ollama on VPS

Once your Ollama instance is running on a VPS with a stable API endpoint, you can integrate it into a wide range of real-world applications. These are the use cases I see most often in production environments.

Private AI Chatbot for Your Team

Use Open WebUI on top of Ollama to give your team a private, self-hosted chat interface. No data ever leaves your server, which makes it suitable for legal, medical, financial, or confidential internal content.

Automated Workflows with n8n

n8n is a popular open-source automation tool. Connect it to your Ollama API to build workflows that summarize emails, extract data from documents, generate content, or reply to support tickets, all without paying per API call.

RAG Pipeline Over Internal Documents

Combine Ollama with a vector database like Chroma or Qdrant to build a retrieval-augmented generation system. Your Ollama model can answer questions based on your internal documents, knowledgebase, or codebase.

API Backend for Your Web App

Because Ollama exposes an OpenAI-compatible REST API, any application that already integrates with OpenAI can be pointed to your Ollama endpoint instead with no code changes. Just update the base URL and eliminate your OpenAI bill.

AI-Powered Dev Tools

Use Ollama with coding-focused models like qwen2.5-coder or deepseek-coder to power local code completion, review automation, or documentation generation inside your CI/CD pipeline.

Quick Reference: All Commands in One Place

Here is every command from this guide grouped by task for easy reference. Bookmark this section so you do not have to scroll through the full guide every time you need a command.

Installation

ssh root@YOUR_VPS_IPapt update && apt upgrade -ycurl -fsSL https://ollama.com/install.sh | shsystemctl status ollama

Model Management

ollama pull llama3.2:3bollama run llama3.2:3bollama listollama rm MODEL_NAME

API Test

curl http://localhost:11434curl http://localhost:11434/api/generate -d '{"model":"llama3.2:3b","prompt":"Hello","stream":false}'

Service Management

sudo systemctl start ollamasudo systemctl stop ollamasudo systemctl restart ollamasudo systemctl enable ollamajournalctl -u ollama -f

Firewall

sudo ufw allow 22/tcpsudo ufw allow 80/tcpsudo ufw allow 443/tcpsudo ufw deny 11434sudo ufw enable

FAQs

These are the questions I get asked most often by people setting up Ollama on a VPS for the first time.

A Final Note To The Reader

You have now set up a fully functional, private AI API on your own VPS. 

I have walked you through every step personally, from SSH access to Nginx configuration and security hardening. Every command in this guide has been tested on live Ubuntu VPS instances.

Your VPS is now running Ollama as a persistent background service with HTTPS access, firewall rules in place, and authentication protecting the endpoint. 

You own the infrastructure, the models, and the data.