Which Mac for IOTA Train at Home? Hardware requirements explained
Verified with T@H 3.7.2 on macOS 26.6.2 (November 15, 2026)
Published November 17, 2026 · 5 min read
If you are wondering whether your Mac can run Train at Home, or which Mac to dedicate to it, the answer is in a short official requirements page. This guide quotes it, adds what we could check on a real Mac, and says plainly where the official documentation stays silent. If you don't yet know what the app does, start with what IOTA Train at Home is.
The official requirements, quoted
The Hardware & OS Requirements page says:
- Platform: "macOS Apple Silicon. M‑class chips are recommended for efficiency but not required." Linux is planned; "Windows is not yet supported."
- Chip: "modern Apple Silicon capable of sustaining training loads."
- Memory: "16 GB RAM minimum."
- Disk: at least 10 GB of available space for binaries, caches and logs, and "up to 30 GB may be required depending on the current model size being trained."
- Network: "Fast, stable internet connection is imperative; training will stall on intermittent links."
- Power: "Keep the machine plugged in and awake to maintain contribution."
For best performance it recommends "Apple Silicon M‑class chips (M1/M2/M3)", "16–32 GB RAM to avoid swapping under load", "Wired Ethernet or strong Wi‑Fi 6", and a cooling pad or elevated stand against thermal throttling.
Chip: Apple Silicon only
Apple Silicon is the only supported Mac platform. On the Mac we checked (an M4 Max, Train at Home 3.7.2), the app's executable contains only an arm64 slice, and the app declares macOS 12.0 as its minimum system version. So an Intel Mac is not a candidate, which matches the official wording. The docs do not give a minimum macOS version, so the 12.0 figure comes from the app bundle, not from documentation, and could change in a later release.
Which M-series generation? The docs name M1, M2 and M3 as examples and do not rank chips or publish any benchmark. We found no official statement about M4 or newer, or about Pro, Max and Ultra variants versus the base chip. "More powerful computers usually contribute more" is the only official guidance, and it is qualitative.
RAM and storage
- RAM. Plan on 16 GB or more. The recommendation to go up to 32 GB is about avoiding swapping while training. The docs state no figure for what the app actually uses, and we did not measure it, so we can't say whether 16 GB leaves room for other work.
- Storage. Keep at least 10 GB of space available, and closer to 30 GB if you can, since the documented upper figure depends on the model size of the current run. That matters most on a small internal drive.
Mac mini or MacBook?
The official docs do not compare form factors. What they do say points in one direction, so the following is our reading of those statements, not an official recommendation:
- Training has to stay on: "Nothing continues running when the app is closed or your computer goes to sleep." A machine that is plugged in and never leaves your desk fits that well.
- The FAQ notes that training can mean "higher fan activity" and "faster battery drain if unplugged", and the user guide says "Keep the laptop powered and on a stable network for steady rewards. Thermal throttling may reduce contribution."
- A laptop you carry around will sleep, change networks and run on battery. A dedicated Mac that stays in place avoids all three.
A MacBook works if it stays on AC power, stays awake and keeps a steady connection; the keep-awake guide covers the sleep side. We don't have measured thermal or throughput comparisons between a Mac mini and a MacBook, so we won't claim one is faster. If you plan to use the same Mac for daily work, remember the FAQ's own caveat: training "can" slow your computer down when you multitask, and you can pause it at any time.
Network
Both the requirements page and the FAQ put stability ahead of speed: "A stable internet connection helps your computer stay connected to the network and send results back." The requirements page also advises avoiding "captive portals or VPNs that frequently re-authenticate." No Mbps threshold is published anywhere we looked. Wired Ethernet is recommended, with Wi‑Fi 6 as the alternative. For ports, P2P and what is normal, see network, ports and P2P.
One more official point is worth knowing before you buy anything: the FAQ says a strong machine and fast internet can still perform poorly, because "your work relies on the work of others" in a swarm. Hardware is one input among several, and neither we nor the docs can promise a result from a given Mac.
What we could not verify
- A minimum macOS version in the documentation (only the app bundle's declared 12.0).
- Real memory use, CPU or GPU load, and bandwidth use of the app by Mac model.
- Any difference between M1, M2, M3, M4 or Pro/Max/Ultra variants, or between Mac mini and MacBook.
- Whether the 10 to 30 GB of disk is the app's own footprint or a working-space figure beyond what the docs state.
Checking a Mac you already own
Open the Apple menu, then About This Mac: the chip must say Apple M-something, with at least 16 GB of memory. For disk, look at the available space in System Settings, General, Storage. Once the app runs, the install guide covers the first launch and the optimal setup guide covers keeping it steady. The subnera menu bar app, an unofficial companion, can then show what a Mac is doing from the official log: download it here. It is independent and not affiliated with IOTA, Macrocosmos or Bittensor.
Sources
Related guides
Want to see this on your own Mac?
subnera shows the real queue position and phase in your menu bar.
Install subnera