What You'll Need

Operating System

Windows 10/11, macOS 11+, or Linux

Memory (RAM)

8GB minimum, 16GB+ recommended

Disk Space

5GB minimum for basic models

GPU (Optional)

NVIDIA GPU with 6GB+ VRAM for faster processing

Installation

Select your operating system:

1

Download Ollama

Visit the official Ollama website and download the Windows installer.

Download for Windows
2

Run the Installer

Double-click the downloaded OllamaSetup.exe file and follow the installation wizard. Accept the default settings.

Note: If Windows SmartScreen appears, click "More info" then "Run anyway". Ollama is safe and open-source.
3

Verify Installation

Open Command Prompt or PowerShell and run:

ollama --version

You should see a version number like ollama version 0.5.x

4

Download Your First Model

Download the recommended model for general AI tasks:

ollama pull llama3.2:3b

This downloads a 2GB model optimized for most tasks. The download may take a few minutes depending on your internet speed.

5

Test the Model

Verify everything works by running a quick test:

ollama run llama3.2:3b "Hello, are you working?"

If you see a response, Ollama is ready! Press Ctrl+C to exit.

Ollama Runs in Background

After installation, Ollama runs automatically as a background service. Look for the llama icon in your system tray (bottom-right of taskbar). Our apps connect to it at localhost:11434.

1

Download Ollama

Visit the official Ollama website and download the macOS app.

Download for macOS
Apple Silicon (M1/M2/M3/M4) Intel Both supported - universal binary
2

Install the App

  1. Open the downloaded Ollama-darwin.zip file
  2. Drag Ollama.app to your Applications folder
  3. Double-click Ollama in Applications to launch it
  4. If prompted, click "Open" to allow the app from an identified developer
3

Allow Background Access

When Ollama first runs, it may ask for permission to run in the background. Click Allow to let it serve AI requests.

You'll see a llama icon appear in your menu bar (top-right).

4

Open Terminal and Download a Model

Open Terminal (search in Spotlight with Cmd+Space) and run:

ollama pull llama3.2:3b

This downloads the recommended 2GB model. Wait for the download to complete.

5

Test the Model

Verify everything works:

ollama run llama3.2:3b "Hello, are you working?"

If you see a response, you're all set! Press Ctrl+C to exit.

Mac Performance Tips

Apple Silicon (M1/M2/M3/M4)

Excellent performance. The GPU is used automatically. 16GB+ unified memory recommended for larger models.

Intel Macs

Good performance on CPU. Stick to smaller models like llama3.2:3b or phi3:mini for best results.

Alternative: Install via Homebrew

If you prefer Homebrew:

brew install ollama ollama serve & ollama pull llama3.2:3b
1

Install with One Command

Open your terminal and run the official install script:

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

This automatically detects your system and installs Ollama.

2

Start the Ollama Service

The installer usually starts Ollama automatically. If not, run:

ollama serve

Or to run as a systemd service:

sudo systemctl enable ollama sudo systemctl start ollama
3

Download a Model

ollama pull llama3.2:3b
4

Test the Model

ollama run llama3.2:3b "Hello, are you working?"

Recommended Models

Select your GPU's VRAM to see which models work best for your hardware. Apple Silicon users: match your unified memory to these tiers.

Integrated GPU, GT 1030, GTX 1050 M1/M2/M3 with 8GB unified (~3-4GB available)
llama3.2:3b Required

Summarization, simplification, study questions

2GB download
phi3:mini Optional

Fast responses, good for quick tasks

2.3GB download

Install Commands

ollama pull llama3.2:3bollama pull phi3:mini
Run one model at a time. Close other GPU-intensive apps for best performance.
RTX 3050, RTX 4060, GTX 1070/1080 M1/M2/M3 with 16GB unified (~11GB available)
llama3.2:3b Required

Summarization, simplification, study questions

2GB download
llava Recommended

Image descriptions and diagram understanding

4.7GB download
mistral:7b Optional

Better writing analysis and text quality

4.1GB download

Install Commands

ollama pull llama3.2:3bollama pull llavaollama pull mistral:7b
Good balance of speed and quality. Most users will be well-served at this tier.
RTX 3060 12GB, RTX 4070, RX 6700 XT M1/M2/M3 Pro with 18GB unified (~13GB available)
llama3.1:8b Recommended

Noticeably better quality for all tasks

4.7GB download
llava Recommended

Image descriptions and diagram understanding

4.7GB download
mistral:7b Optional

Creative writing and detailed analysis

4.1GB download

Install Commands

ollama pull llama3.1:8bollama pull llavaollama pull mistral:7b
Use llama3.1:8b as your default instead of 3b for significantly better results.
RTX 4080, RTX 3070 Ti, RX 7800 XT M1/M2/M3 Pro/Max with 24GB unified (~19GB available)
llama3.1:8b Recommended

High quality default for all tasks

4.7GB download
llava:13b Recommended

Superior image understanding and descriptions

8GB download
gemma2:9b Optional

Strong reasoning and comprehension

5.4GB download

Install Commands

ollama pull llama3.1:8bollama pull llava:13bollama pull gemma2:9b
Can keep multiple models loaded simultaneously for instant switching.
RTX 3090, RTX 4090, RTX A5000 M2/M3 Max with 32GB unified (~27GB available)
llama3.1:13b Recommended

Excellent quality for all tasks

7.4GB download
llava:13b Recommended

Superior image understanding

8GB download
deepseek-r1:14b Optional

Advanced reasoning and problem solving

8.5GB download

Install Commands

ollama pull llama3.1:13bollama pull llava:13bollama pull deepseek-r1:14b
Run larger models for significantly better output quality. Plenty of headroom.
Multi-GPU, RTX A6000, Professional GPUs M2/M3 Ultra with 64GB+ unified (~59GB+ available)
mixtral:8x7b Recommended

State-of-the-art mixture-of-experts model

26GB download
llama3.1:13b Recommended

High quality, fast responses

7.4GB download
llava:34b Optional

Best-in-class image understanding

20GB download

Install Commands

ollama pull mixtral:8x7bollama pull llama3.1:13bollama pull llava:34b
Maximum quality. Can run the largest models with room to spare.
Not sure about your VRAM?

Windows: Open Task Manager → Performance → GPU to see dedicated GPU memory. Mac: Apple menu → About This Mac shows your total unified memory—subtract 4-8GB for macOS and apps to estimate what's available for models.

Verify Connection

Fiavaion apps automatically detect Ollama. Here's how to verify it's running:

Check in Browser

Open localhost:11434 in your browser. You should see "Ollama is running".

Check in Terminal

Run ollama list to see your installed models.

Check in Our Apps

Open any Fiavaion app with AI features. It will show a green indicator if Ollama is connected.

Troubleshooting

"Ollama not detected" error
  1. Make sure Ollama is running (check for the llama icon in system tray/menu bar)
  2. Try restarting Ollama
  3. Verify localhost:11434 is accessible
  4. Check if a firewall is blocking the connection
Slow responses
  • Try a smaller model: ollama pull phi3:mini
  • Close other memory-intensive applications
  • If using CPU only, responses will be slower than with a GPU
  • Check available RAM with Task Manager (Windows) or Activity Monitor (Mac)
Model download fails
  • Check your internet connection
  • Ensure you have enough disk space (models are 2-8GB each)
  • Try a smaller model first: ollama pull phi3:mini
  • If behind a proxy, configure it in your terminal environment
Out of memory errors
  • Use a smaller model (3b or mini variants)
  • Close other applications to free up RAM
  • Restart Ollama to clear any stuck processes
  • Consider upgrading RAM if you frequently hit limits
macOS: "Cannot be opened" error
  1. Right-click (or Ctrl+click) the Ollama app
  2. Select "Open" from the context menu
  3. Click "Open" in the dialog that appears
  4. This only needs to be done once
Windows: Ollama not starting
  1. Check Windows Services (services.msc) for "Ollama"
  2. Right-click and select "Start" if it's stopped
  3. Try reinstalling Ollama
  4. Run as Administrator if needed

Your Data Stays Local

With Ollama, all AI processing happens on your computer. Your documents, text, and images are never sent to any server. This makes it safe for:

  • Student work (FERPA compliant)
  • Healthcare data (HIPAA safe)
  • Personal documents (GDPR friendly)
  • Sensitive business information

Need More Help?

Browse All Models Ollama GitHub AI Features Guide
Back to Documentation