Install Ollama on Every OS

System Requirements — What Hardware Do You Need?

Before installing, check if your machine meets the requirements. Ollama doesn't need a powerful machine — a regular laptop with 8 GB RAM can run 7-8B models. Having a GPU significantly speeds things up

Spec	Minimum	Recommended	Enterprise
OS	macOS 11+, Windows 10+, Linux	macOS 14+, Windows 11, Ubuntu 22.04+	Ubuntu 22.04 LTS Server
RAM	8 GB	16 GB	64-128 GB
Disk Space	10 GB	50 GB	500 GB+ SSD
CPU	x86_64 or Apple Silicon	Apple M1+ or Intel i7+	AMD EPYC / Intel Xeon
GPU (Optional)	Not required (CPU works)	NVIDIA RTX 3060+ (VRAM 8 GB+)	NVIDIA RTX 4090 / A100 (VRAM 24-80 GB)
Models You Can Run	7-8B (Llama 3.1 8B, Gemma 2 9B)	14-34B (Phi-4, CodeLlama 34B)	70B+ (Llama 3.1 70B, Qwen 72B)

Install on macOS

macOS was the first platform Ollama supported and works best, especially on Apple Silicon:

Method 1: Download from Website (Recommended)

1. Go to ollama.com and click "Download for macOS"
2. Open the downloaded Ollama-darwin.zip file
3. Drag Ollama.app to Applications
4. Open Ollama.app — you'll see the Llama icon in the Menu Bar (top right)
5. Open Terminal and type ollama run llama3.1

Method 2: Via Homebrew

For developers who already use Homebrew:

brew install ollama
ollama serve

Install on Windows

Ollama supports Windows 10 and above (64-bit), including NVIDIA GPU, AMD GPU (Radeon RX 6000+), and CPU-only:

1. Go to ollama.com and click "Download for Windows"
2. Run OllamaSetup.exe
3. Follow the installation steps (click Next → Install → Finish)
4. Ollama will run as a Background Service automatically — look for the icon in System Tray
5. Open PowerShell or Command Prompt and type ollama run llama3.1

Install on Linux

Linux is the best platform for setting up an organization's Ollama Server — stable and resource-efficient:

Method 1: Automatic Script (Recommended)

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

This script downloads Ollama, installs it, creates a systemd service, and starts the Ollama Server automatically. Supports Ubuntu, Debian, Fedora, CentOS, RHEL, Arch Linux, and more

Method 2: Download Binary Manually

# Download Binary
curl -L https://ollama.com/download/ollama-linux-amd64 -o /usr/local/bin/ollama
chmod +x /usr/local/bin/ollama

# Start Server
ollama serve

Method 3: Docker

For organizations already using Docker — ideal for enterprise-level infrastructure management:

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

# NVIDIA GPU
docker run -d --gpus=all -v ollama:/root/.ollama -p 11434:11434 --name ollama ollama/ollama

# Run Model
docker exec -it ollama ollama run llama3.1

Run Your First AI Model

After installation, regardless of OS, the next step is the same — open Terminal and type:

ollama run llama3.1

The first time, Ollama will download the model (~4.7 GB for Llama 3.1 8B), which may take 2-10 minutes depending on your internet speed. After downloading, you'll see a prompt to type your question:

>>> Hello, what AI are you?

Hello! I am Llama 3.1, a Large Language Model developed by Meta.
I can help answer questions, write articles, translate languages, and more.
How can I help you?

Done! You now have a private AI chatbot on your own machine — everything you type stays on your computer, nothing is sent outside. Press /bye to exit, or Ctrl+D

Essential Commands You Need to Know

After installing Ollama, these are the commands you'll use most frequently:

Command	What It Does	Example
ollama run <model>	Run a model (auto-downloads if not present)	ollama run llama3.1
ollama pull <model>	Download a model without running it	ollama pull qwen2.5:72b
ollama list	List downloaded models	ollama list
ollama rm <model>	Remove a model from your machine	ollama rm mistral
ollama show <model>	Show model info (size, parameters)	ollama show llama3.1
ollama ps	Show currently running models	ollama ps
ollama serve	Start Ollama Server (usually runs automatically)	ollama serve
ollama cp <src> <dst>	Copy a model to create a new one from existing	ollama cp llama3.1 my-assistant

Test — Run Multiple Models for Comparison

After installation, try running multiple models to compare — each has different strengths:

# All-round model (by Meta)
ollama run llama3.1

# Strong Asian languages (by Alibaba)
ollama run qwen2.5

# Reasoning model - thinks before answering (by DeepSeek)
ollama run deepseek-r1:8b

# Coding model (by Alibaba)
ollama run qwen2.5-coder

# Google model — very fast
ollama run gemma2

# Vision model — reads images
ollama run llava

Try asking the same question to multiple models and compare answers — you'll see each has a different response style. Some excel at Asian languages, others at coding. For detailed model comparisons, see EP.1 Popular Models Table

Install GUI — Open WebUI (No Terminal Needed)

For those who prefer not to type commands in Terminal, install Open WebUI — a beautiful ChatGPT-like web interface that connects to Ollama instantly:

# Install Open WebUI with Docker (Docker required)
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 your browser to http://localhost:3000 — create an admin account, select a downloaded model, and start using it! Open WebUI is perfect for organizations that want all employees to use AI without opening Terminal — just open the web and type like ChatGPT

Troubleshooting Common Issues

Problem	Cause	Solution
"Error: model requires more memory"	Not enough RAM for the selected model	Use a smaller model, e.g., switch from 70B to 8B, or use a quantized version ollama run llama3.1:8b-q4_0
"Error: listen tcp 127.0.0.1:11434: bind: address already in use"	Ollama is already running (possibly in background)	Stop the running Ollama first: pkill ollama (Linux/macOS) or close from System Tray (Windows)
GPU not being used (very slow)	Drivers not updated or Ollama can't detect GPU	Update NVIDIA drivers to latest version. Verify with nvidia-smi then restart Ollama
Poor Thai language / nonsensical answers	Some models aren't good at Thai	Switch to qwen2.5 or llama3.1:70b which has better Thai support
Model download slow / interrupted	Slow or unstable internet	Run ollama pull <model> again — it will resume from where it stopped
Can't access from other machines (Server)	Ollama listens on localhost only	Set environment variable: OLLAMA_HOST=0.0.0.0 ollama serve
Disk full	Downloaded too many models	Remove unused models: ollama rm <model> List models with ollama list

For Organizations — Set Up Ollama Server for Everyone

To let everyone in the organization use Ollama together without installing on every machine — just set up one server and let everyone access via Open WebUI:

# 1. Install Ollama on Linux server
curl -fsSL https://ollama.com/install.sh | sh

# 2. Configure to listen on all IPs (not just localhost)
sudo systemctl edit ollama
# Add this line:
# [Service]
# Environment="OLLAMA_HOST=0.0.0.0"

# 3. Restart Ollama
sudo systemctl restart ollama

# 4. Download desired models
ollama pull llama3.1
ollama pull qwen2.5
ollama pull deepseek-r1:8b

# 5. Install Open WebUI with Docker
docker run -d -p 3000:8080 \
  -e OLLAMA_BASE_URL=http://host.docker.internal:11434 \
  -v open-webui:/app/backend/data \
  --name open-webui \
  --restart always \
  ghcr.io/open-webui/open-webui:main

Once done, tell employees to open http://<server-ip>:3000 in their browser — everyone gets a ChatGPT-like AI chatbot instantly. All data stays within the organization's network, compliant with PDPA requirements and the organization's data security policies

Ollama Series — Read More

"Installing Ollama takes just 5 minutes, but what you get is a private AI that works 24/7 without monthly subscription fees."

- Saeree ERP Team

References

• Ollama — Download Page (Official)
• Ollama — GitHub README (Installation Guide)
• Ollama — GPU Compatibility Guide
• Open WebUI — GitHub (GUI for Ollama)