Always-on AI coding tools
Keep Claude Code, Codex CLI, Gemini CLI and similar tools running in tmux or screen, editing code and running tests for hours while you check in over SSH.
AI workstation · Online 24/7
A dedicated Mac mini that runs your AI tools 24/7
Keep AI coding tools, AI agents and local LLMs running on your own remote Mac mini. Shut down your laptop, lose Wi-Fi or close the lid, and the work keeps going.
A dedicated Mac mini: M4 from US$160/month, M6 at US$280/month, ready in 15–30 minutes after payment.
You log off, your AI keeps working
Close your laptop and the AI jobs on it stop. On a remote Mac they can run from morning to night, and the results are waiting for you the next day.
Your laptop
Lumio remote Mac
00:0006:0012:0018:0024:00
On, jobs runningOff, jobs stopped
Why run AI on a remote Mac
Many AI jobs run for hours. On your own laptop they hog resources and stop when you leave. An always-on dedicated Mac is a better home for them.
AI agents that use Mac apps
AI agents can work on a full macOS desktop, driving the browser, Xcode, Finder and other Mac apps, something a Linux VPS cannot offer.
Run open models locally
Apple silicon shares unified memory with the GPU, so Ollama, LM Studio or MLX can run open models like Llama, Qwen and Gemma right on the machine.
Keep your own computer free
Hand long inference runs, batch jobs and automation to the remote Mac. Your laptop stays responsive, and you can check results from your phone.
AI tools you can run on this Mac
You get a complete physical macOS machine with root access. Install anything, just like on your own Mac.
Use your own AI service accounts, following each provider's terms and supported regions
AI coding tools
- Claude Code
- Codex CLI
- Gemini CLI
- Cursor
- VS Code
AI desktop apps
- ChatGPT
- Claude
Local models
- Ollama
- LM Studio
- MLX
- llama.cpp
Automation
- Shortcuts
- AppleScript
- Python
- Node.js
Developer stack
- Homebrew
- Xcode
- Docker
- Git
- tmux
How large a local model fits in 16GB
M4 and M6 plans both come with 16GB of unified memory. Here is a rough guide for common open models at 4-bit quantization. Actual usage varies by model and context length.
If you mainly use cloud AI services such as Claude Code or Codex CLI, the model runs on the provider's side and needs very little local memory, so M4 is plenty. The larger your local models, the more the newer M6 chip pays off.
M4 or M6 for AI?
Both are dedicated physical machines with 16GB of memory. The difference is chip speed and price.
Mac mini M4
US$160/month
M4 for AI coding tools and agents
If you mostly use cloud AI tools like Claude Code and Codex CLI, or run small models around 7B, M4 is plenty, from US$160/month.
Explore Mac mini M4 -> Mac mini M6 US$280/monthM6 for faster local inference
If you often run open models locally, do batch inference or run several agents at once, the 12-core M6 and its stronger AI performance fit better, at US$280/month.
Explore Mac mini M6 ->Mac mini for AI FAQ
Common questions about running AI on a remote Mac.
Can I use ChatGPT, Claude or Gemini on this Mac?
You can install and run these services' official apps and command-line tools. You need your own account and must use each service in its supported countries and regions and under its terms. Lumio provides the Mac hardware and network only and does not provide or resell accounts for any AI service.
Will my AI jobs stop when my computer is off?
No. The jobs run on the remote Mac, independent of your own computer. Run command-line tools inside tmux or screen so they keep going after SSH disconnects, and reattach to check progress.
What is the largest model I can run locally?
16GB of unified memory suits 7B–8B models at 4-bit quantization. 12B–14B models run but are tight. Models of 30B and above exceed the memory and are not recommended locally.
Is there an NVIDIA GPU? Can I use CUDA?
No. The Mac mini uses the GPU and Neural Engine built into Apple silicon and does not support CUDA. Run local models with Apple silicon tools such as Ollama, LM Studio, MLX and llama.cpp.
Are my code and data safe?
Each Mac mini is delivered to a single customer and never shared, and we never audit or retain user data. Once your plan expires and you confirm you will not renew, the machine is automatically erased within 24 hours, permanently removing your code, tokens and account data.
Can I run several AI agents at once?
Yes, depending on CPU and memory load. Agents that call cloud AI services use little and several can run together. If each one runs a local model, the 16GB of memory becomes the limit and M6 is the better choice.
Give your AI a Mac that never sleeps
A dedicated Mac mini from US$160/month, ready in 15–30 minutes after payment and ready for VNC and SSH.