The M6 Mac mini is on its way to running Omarchy. Joshua Warren of the Omarchy M team posted on October 18: “Installed a bootloader, got the USB debug console working. It runs the second-stage loader, and the Linux kernel and initramfs are loaded into RAM.”
How fast this is moving
- The mini arrived October 16. Bring-up started October 17 with the goal of going “from macOS About screen to Omarchy About screen.”
- Next up: U-Boot handoff and a Linux shell, then HDMI and the Omarchy desktop, then the Apple Neural Engine and GPU.
- He’s putting AI agents on the job to speed up the bring-up.
The Neural Engine angle
Warren identified the M6’s ANE as a new target, H18G: two 16-core engines whose firmware iBoot loads before macOS starts. He compiled the first H18G program the same day. His omarchy-mlx project already runs Vulkan and ANE backends on Apple Silicon under Linux, so a working ANE on M6 slots into an existing stack.
Benchmarks (on macOS, before bring-up)
Against his M1 Max, the M6 was 3.1x faster at prefill (1,046 vs 339 tokens/sec). On decode, Qwen3.6-35B-A3B ran at 69.4 vs 75.6 tokens/sec, “92% of the Max’s performance with only 42.5% of the bandwidth.” A 27B dense model hit the memory wall at 9.6 vs 19.5.
My take
Linux on a brand-new Apple chip within days of delivery would have been unthinkable a few years ago. This builds on the Muse Spark Open cluster run and the Omarchy for Mac image.
