mac-mini-ai（mac 迷你 ai）

Mac Mini AI — The $599 AI Node

The Mac Mini is the most cost-effective hardware for local AI. Starting at $599 with 16GB of unified memory, it runs 7B-14B models comfortably. Stack three Mac Minis for the cost of one month of cloud GPU rental — and they run forever with zero ongoing costs.

This skill turns one Mac Mini into an AI server and multiple Mac Minis into a fleet.

Mac Mini configurations for AI

The Mac Mini fleet strategy

Three Mac Minis (32GB each) for $3,000 give you:

• 96GB total unified memory across the fleet
• Each runs a different model simultaneously
• The router picks the best device for every request
• $0/month after purchase — no cloud API costs

Mac Mini #1 (32GB) — llama3.3:70b (quantized)  ─┐
Mac Mini #2 (32GB) — codestral + phi4            ├──→  Router  ←──  Your apps
Mac Mini #3 (32GB) — qwen3:14b + embeddings     ─┘

Setup

pip install ollama-herd    # PyPI: https://pypi.org/project/ollama-herd/

On one Mac Mini (the router):

herd

On every other Mac Mini:

herd-node

Devices discover each other automatically. No IP configuration, no Docker, no Kubernetes.

Use your Mac Mini

Chat with an LLM

from openai import OpenAI

client = OpenAI(base_url="http://localhost:11435/v1", api_key="not-needed")
response = client.chat.completions.create(
    model="phi4",
    messages=[{"role": "user", "content": "Write a Python web scraper"}],
    stream=True,
)
for chunk in response:
    print(chunk.choices[0].delta.content or "", end="")

Ollama API

curl http://localhost:11435/api/chat -d '{
  "model": "gemma3:12b",
  "messages": [{"role": "user", "content": "Explain recursion simply"}],
  "stream": false
}'

Image generation (optional)

uv tool install mflux    # Install on any Mac Mini
curl -o art.png http://localhost:11435/api/generate-image \
  -H "Content-Type: application/json" \
  -d '{"model": "z-image-turbo", "prompt": "a stack of Mac Minis glowing", "width": 512, "height": 512}'

Speech-to-text

curl http://localhost:11435/api/transcribe -F "file=@meeting.wav" -F "model=qwen3-asr"

Embeddings for RAG

curl http://localhost:11435/api/embed \
  -d '{"model": "nomic-embed-text", "input": "Mac Mini home server local AI"}'

Best models for Mac Mini

Monitor your Mac Mini fleet

Dashboard at http://localhost:11435/dashboard — see every Mac Mini's status, loaded models, and queue depths.

# Fleet overview
curl -s http://localhost:11435/fleet/status | python3 -m json.tool

# Model recommendations for your hardware
curl -s http://localhost:11435/dashboard/api/recommendations | python3 -m json.tool

Works with any OpenAI-compatible tool

Full documentation

• Agent Setup Guide
• API Reference
• Configuration

Contribute

Ollama Herd is open source (MIT). Built for the Mac Mini fleet community:

• Star on GitHub — help other Mac Mini owners find us
• Open an issue — share your Mac Mini fleet setup
• PRs welcome from humans and AI agents. CLAUDE.md gives full context.
• Running a Mac Mini cluster? We'd love to hear about it.

Guardrails

• No automatic downloads — model pulls require explicit user confirmation.
• Model deletion requires explicit user confirmation.
• All requests stay local — no data leaves your network.
• Never delete or modify files in ~/.fleet-manager/.