Run powerful AI models completely offline
Want ChatGPT-style AI that runs 100% on your computer?
Ollama lets you use powerful language models completely offline—with better privacy, customization, and no subscription fees!
This simple, step-by-step guide covers:
- ✅ Why run AI locally? (Privacy, customization & offline use)
- ✅ How to install Ollama (Windows/Mac/Linux)
- ✅ Downloading & running models (Like Llama 3, Mistral, and more)
- ✅ Using Ollama with ComfyUI (For better Stable Diffusion prompts!)
Why Run LLMs Locally with Ollama?
1. Total Data Privacy
- No cloud dependency – Your chats, prompts, and data stay on your device.
- Crucial for sensitive work (legal, medical, or proprietary projects).
2. Full Offline Access
- Once downloaded, models work without internet—great for travel or remote work.
3. Customization Freedom
- Modify models for specific tasks (coding, creative writing, etc.).
- Avoid API limits or censorship from services like ChatGPT.
What is Ollama?
Ollama is a free, open-source tool that lets you:
- Download & run popular AI models (Llama 3, Mistral, Phi, etc.)
- Use them via command line or apps like ComfyUI
- Works on Windows, Mac, and Linux
Part 1: Installing & Running Ollama
Step 1: Download Ollama
- Go to ollama.com
- Click Download for your OS (Windows/macOS/Linux).
- Run the installer (no special settings needed).
✅ Check it's running:
Windows: Look for the Ollama icon in the system tray.
Mac/Linux: Open Terminal and type ollama --version.
Step 2: Stop Ollama from Auto-Starting (Optional)
To save resources, disable auto-start:
Windows:
- Press Ctrl + Shift + Esc → Startup tab → Disable Ollama.
Mac/Linux:
systemctl disable ollama
Step 3: Choose Where to Save Models (For Big Downloads)
By default, models save to your C: drive. To change this:
- Open Environment Variables (Windows: Search "ENV" → "Edit system environment variables").
- Under User variables, click New.
- Add:
Name:OLLAMA_MODELS
Value:D:\AI\Models(or any folder with space!)
*(Now 20GB+ models won't fill your main drive!)*
Step 4: Download & Run Your First Model
Open Command Prompt (Windows) or Terminal (Mac/Linux).
A. Pull a Model
ollama pull llama3
(This downloads Meta's Llama 3—great for general use.)
B. Run It!
ollama run llama3
Now chat directly in your terminal—no internet needed!
🔹 Other Top Models:
| Model | Best For | Command |
|---|---|---|
| mistral | Fast, lightweight | ollama run mistral |
| codellama | Programming help | ollama run codellama |
| llava | Image descriptions | ollama run llava |
(See all models at Ollama's library.)
Step 5: Ollama CLI Cheat Sheet
| Command | What It Does |
|---|---|
ollama list |
Shows installed models |
ollama pull <model> |
Downloads a new model |
ollama run <model> |
Starts chatting with the model |
ollama serve |
Runs Ollama as a local API server |
(Full CLI reference: Ollama's GitHub.)
⚠️ Model too big? Try phi3-mini or tinyllama for low-RAM PCs.
Ollama is a game-changer for local AI—whether you're refining SD prompts, coding offline, or keeping data private.
Ready to try it? Start with llama3 and explore!