One physical Mac mini per customer, billed daily, priced in USD. The whole machine is yours: full admin rights, VNC desktop plus SSH access, and a dedicated IPv4. Built as a reliable base for Xcode cloud builds, macOS CI runners, and local MLX inference — credentials delivered by email within 10 minutes of payment.
$ ssh dev@tokyo-07.armmini.com
Mac mini · M4 Pro · 64GB · macOS 15
$ xcodebuild -scheme App archive
** ARCHIVE SUCCEEDED ** 4m 12sNot a slice of a shared VM — a whole Mac mini, handed to you. Machines live in a data center rack, on dual power feeds, running 365 days a year.
One full machine per customer — no sharing CPU, RAM, or disk with anyone else. There's no hypervisor layer; what you see in sysctl is the actual box. You get a full admin account from day one, free to install any Xcode version, Homebrew, or system extensions. We keep no backdoor accounts.
M4 / M4 Pro running natively on physical hardware — GPU and Neural Engine fully passed through, zero virtualization overhead. Build times don't drift: a full archive takes 4 minutes today, 4 minutes next month. Your CI duration graph stays flat.
Rent for as little as one day, priced in USD — the longer the term, the lower the rate (a week ≈ 2.7× the daily rate, a month = 5× the daily rate). Rent for three days to validate a build before release, or keep a runner on for a whole quarter. No long-term lock-in, and nothing renews automatically once your term ends.
Your machine arrives as a clean macOS install with an Xcode version of your choice pre-installed — the commands below just work. We'll also tell you when we're not the right fit: one-off one-hour builds, or anything that needs Windows — skip us.
Full builds, archiving, and uploads to App Store Connect — you fully control certificates and keychains, and can pull any private dependency. On the M4 tier, full builds of large Swift projects typically run more than twice as fast as on a local M1.
$ xcodebuild -scheme App \
-destination generic/platform=iOS \
archive
$ xcrun altool --upload-app ...
UPLOAD SUCCEEDEDRegister a self-hosted runner for GitHub Actions or GitLab, on a monthly plan. DerivedData and simulator caches stay on local disk, so incremental builds skip the cold start every time.
$ ./config.sh --url "$REPO_URL" \
--labels macos,arm64
$ ./svc.sh install && ./svc.sh start
√ runner connectedApple Silicon unified memory feeds straight into your model — the 64GB tier can load 70B-class quantized weights for inference experiments, with data staying on your own machine.
$ pip install mlx-lm
$ mlx_lm.generate \
--model Qwen2.5-72B-4bit \
--prompt "hello"
tokens/s: 11.2Full build takes 12 minutes on my local M1, 4 minutes here. I moved all my release packaging to the Tokyo node.
Independent iOS developer
The runner's been up for three months without a restart, running 40+ jobs a day. The build-time graph is basically a flat line.
CI lead at a tooling team
64GB of unified memory is just right for a 70B quantized model — quieter and cheaper than renting a GPU instance, and it doubles as my dev machine.
AI hobbyist
Machines live in data center racks, not in anyone's home office. Custom trays, dual power feeds, out-of-band power control — covered for both power and network outages, running 365 days a year. Alongside the three featured below, Singapore and Hong Kong nodes are also in service and can be picked at checkout.
Every plan is a fully dedicated machine, not a VM — includes admin rights, VNC + SSH access, a dedicated IPv4, one free reinstall per billing cycle, and fair-use unmetered bandwidth.
| Plan | Chip | RAM | SSD | Daily | Monthly | Nodes | |
|---|---|---|---|---|---|---|---|
| Mini Basic | M4 | 16GB | 256GB | $21.5 | $107.7 | SG / JP / KR / HK / US-W | Choose |
| Mini StandardPopular | M4 | 24GB | 512GB | $40.1 | $200.7 | SG / JP / KR / HK / US-W | Choose |
| Mini Pro | M4 Pro | 64GB | 2TB | $59.3 | $296.7 | SG / JP / KR / HK / US-W | Choose |
Machines are delivered within 10 minutes of payment in most cases, and within 24 hours at peak. Credentials are sent via encrypted email — have your SSH public key ready before you order.
Match it to your use case: Basic for a single CI project, Standard for daily development, Pro for Unity/UE packaging and large-model inference. If you're not sure, start with Standard — you can upgrade anytime and just pay the difference.
Asia users should default to Singapore, Hong Kong, Tokyo, or Seoul; North America users to Silicon Valley. Check the latency benchmarks before ordering — don't guess your routing.
Choose your billing term in the portal — USDT-TRC20 or Visa / Mastercard / Amex via Stripe, billed in USD. Unpaid orders are automatically canceled after 30 minutes.
Usually within 10 minutes, your VNC/SSH credentials and connection guide arrive by encrypted email — connect and you're ready to run xcodebuild.
Machines are delivered in 10 minutes, from $21.5/day. Browse the plans to find your fit, or head straight to the portal and pick a machine. Questions? support@armmini.com and the console ticket system are always open.
$ rent --mac-mini --tokyo
plan: Mini Standard · M4 · 24GB
billing: $40.1/day · $200.7/mo
✓ credentials sent · eta 10 min