Local Server Setup Guide

Run 70B+ AI models on your Mac and chat from your iPhone — everything stays on your network.

iPhone OwnPodAI

→

WiFi Same network

→

Mac / PC Ollama

→

AI Model 7B — 70B+

0Requirements

OwnPodAI Local Server works with any computer running Ollama — MacBook, Mac Mini, Mac Studio, Linux PC, or Windows PC. The only requirement: same WiFi network as your iPhone.

No special hardware needed. Got an 8GB MacBook Air? You can run 3B models. Got a 64GB Mac Mini? Run 70B models. OwnPodAI adapts to whatever you have.

What can your machine run?

Device Example	RAM	Chip	Models	Quality
MacBook Air M1	8 GB	M1	Llama 3.2 3B, Phi-4 Mini, Qwen 3 4B	Basic
MacBook Air M2/M3	16 GB	M2/M3	Llama 3.2 8B, Gemma 2 9B, Qwen 3 8B, Mistral 7B	Good
MacBook Pro M3/M4	18 GB	M3/M4	Same as 16GB + slightly faster generation	Good
MacBook Pro M3 Pro/M4 Pro	36 GB	M3/M4 Pro	Qwen 2.5 14B, Gemma 2 27B, CodeLlama 13B	Great
Mac Mini M4	32 GB	M4	Llama 13B, Qwen 14B, Gemma 27B, Mistral 22B	Great
Mac Mini M4 Pro	48 GB	M4 Pro	Llama 3.1 70B (Q3), Qwen 3 32B, Command R 35B	Excellent
Mac Mini M4 Pro / Mac Studio	64 GB	M4 Pro/Max	Llama 3.1 70B, Qwen 3 72B, DeepSeek V3, Mixtral 8x7B	Excellent
Mac Studio M2 Ultra	128 GB	M2 Ultra	Llama 3.1 405B (Q4), any model at high quality	Maximum
Mac Pro / Mac Studio M4 Ultra	192 GB	M4 Ultra	Every model at full precision, multiple models simultaneously	Maximum

PC Users: Linux PCs with NVIDIA GPUs (RTX 3060 12GB, RTX 4090 24GB) and Windows PCs via WSL2 also work. GPU VRAM is used for inference — a 24GB RTX 4090 can run most 13B models at full speed.

1Install Ollama

macOS

Linux

Windows

Open Terminal and run:

Terminal

# Install Ollama (one command)
curl -fsSL https://ollama.com/install.sh | sh

Or download from ollama.com — drag to Applications, done.

Verify installation:

Terminal

ollama --version
# Should show: ollama version 0.x.x

Terminal

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

For NVIDIA GPU support, make sure CUDA drivers are installed first.

Download the installer from ollama.com/download/windows and run it. Ollama runs inside WSL2 automatically.

2Download a Model

Pull any model with one command. Here are the best options:

Terminal

# Recommended for 16GB Mac
ollama pull llama3.2

# Best for 32GB Mac
ollama pull qwen2.5:14b

# Best for 64GB Mac (GPT-4 quality)
ollama pull llama3.1:70b

# Qwen 3 72B (best open model as of 2025)
ollama pull qwen3:72b

# Vision model — can understand images
ollama pull llava:13b

Tip: Start with llama3.2 to test your setup. It downloads fast and runs on any Mac. Upgrade to bigger models once everything works.

3Start Ollama for Network Access

By default, Ollama only listens on localhost. To access it from your iPhone, start it with network access:

Terminal

# Allow connections from your local network
OLLAMA_HOST=0.0.0.0 ollama serve

Keep this Terminal window open. Ollama is now listening on port 11434.

Important: This only opens Ollama to devices on your local WiFi network. It does NOT expose anything to the internet. Your router's firewall keeps everything private.

Make it permanent (optional)

To auto-start Ollama with network access every time your Mac boots:

Terminal

# Add to your shell profile
echo 'export OLLAMA_HOST=0.0.0.0' >> ~/.zshrc
source ~/.zshrc

4Find Your Mac's IP Address

You need your Mac's local IP to connect from the iPhone:

Terminal

# Get your local IP
ipconfig getifaddr en0

# Example output: 192.168.1.XXX

Or go to System Settings → WiFi → Details → IP Address.

Quick test

Verify Ollama is accessible from your network:

Terminal

# Test from your Mac (should return "Ollama is running")
curl http://192.168.1.XXX:11434

# Test a model response
curl http://192.168.1.XXX:11434/api/generate -d '{"model":"llama3.2","prompt":"Hello"}'

5Connect OwnPodAI

On your iPhone:

Open OwnPodAI
Go to Settings (gear icon)
Tap Local Server
Enter your Mac's IP: 192.168.1.XXX
Port: 11434
Tap Connect — you should see a green checkmark
Select any model from the list — start chatting

Done! Your iPhone is now chatting with the AI model running on your Mac. All traffic stays on your home WiFi — nothing goes to the internet.

?Troubleshooting

Can't connect from iPhone

Make sure both devices are on the same WiFi network
Check that Ollama is running with OLLAMA_HOST=0.0.0.0
Try pinging your Mac from iPhone's browser: type http://192.168.1.XXX:11434 in Safari
Check macOS Firewall: System Settings → Network → Firewall — allow incoming connections for Ollama

Model is slow

Check available RAM: the model should fit entirely in memory
Close other heavy apps (Chrome, Xcode, etc.)
Try a smaller quantization: ollama pull llama3.1:70b-q4_0
Monitor with: ollama ps to see loaded models

Model runs out of memory

Unload unused models: ollama stop modelname
Use a smaller model — check the RAM table above
Try quantized versions: Q4 uses ~50% less RAM than Q8

→Useful Ollama Commands

Terminal

# List downloaded models
ollama list

# See running models and memory usage
ollama ps

# Remove a model to free space
ollama rm llama3.1:70b

# Update a model to latest version
ollama pull llama3.2

# Stop all running models
ollama stop llama3.2

# Run a quick test in terminal
ollama run llama3.2 "What is 2+2?"

# Check Ollama version
ollama --version

🔒Security Notes

Your data stays private:
• Ollama runs 100% locally — no cloud, no API keys, no accounts
• WiFi traffic between iPhone and Mac never leaves your router
• No ports are exposed to the internet
• Your conversations are not logged by Ollama by default
• For extra security, use a dedicated VLAN or guest network

Do not expose Ollama to the public internet. The OLLAMA_HOST=0.0.0.0 setting is safe on home WiFi but should never be used on public networks or with port forwarding.