Omarchy on the M6 Mac mini now runs on the GPU, and the Neural Engine is powered on
Within days of the first Omarchy boot on Apple’s newest Mac mini, software rendering is gone and the Neural Engine’s control processor is answering. Next target: running inference on it.
Image: @JoshuaSWarren (X)
The GPU is up. “Linux on the M6 Mac Mini, Day 6 – GPU is up!” Joshua Warren of the Omarchy M team posted on X on October 22, 2026. “Many of you noticed in yesterday’s screenshot it used software rendering. I’m excited to share Omarchy on the M6 now utilizes the GPU.”
That closes one of the two big gaps from our report on day five, when Omarchy had just been installed on the Mac mini’s internal drive but the About screen still showed llvmpipe, Mesa’s software renderer. The screenshot attached to the day six post shows the same machine, an Apple Mac mini (M6, J873), now with its graphics line reading “Apple M6 (G18G B1)” instead.
Warren was quick to share the credit. “Miguel has led the way on the M6 bring-up. I definitely recommend following him and his GPU work,” he wrote the same evening, pointing to Miguel C (@Acelogic_). Miguel had his own update: “Omarchy M6 GPU support incoming, Open GL 4.6 and Vulkan 1.4 conformance is in the works,” he posted on October 22, 2026, adding that he had spent a good deal of time over the past month getting the conformance (CTS) tests to pass. Conformance is still a work in progress, not a done deal.
In the day six post, Warren also wrote that “almost all the drivers are done” and that he would be “starting on the Apple Neural Engine bring-up soon.” Soon meant the very next update.
“Neural Engine: powered on, control processor answers,” Warren reported in his day seven update, posted early on October 24, 2026. On the machine learning side, he wrote that MLX “sees the M6 GPU and 15.6 GB of memory”, but that “the first op failed on one GPU helper missing from our driver. Fix is building.”
His stated goal: “Will be running inference on it by morning.” As of that day seven post, inference is a target, not a result.
The Neural Engine driver work happens in the open in Warren’s omarchy-ane repository. Two pull requests merged there on October 24, 2026 add the first handshake with the Neural Engine firmware to the ane_h18 module. The follow-up, PR #196, notes that its own changes have not run on the board yet and that “the M6 run of #195” is separate, so the code is moving as fast as the posts.
This is a community effort by the Omarchy M team, not official M6 support you can install today. Ethernet, which was still on the to-do list on day five, is not mentioned in either update. But in a week, the M6 Mac mini has gone from the macOS About screen to Omarchy with a working GPU and a Neural Engine that answers, and we will keep following it.
Source: https://x.com/JoshuaSWarren/status/2113343313768067543
