mlx-omarchy v0.5.0 released with a claimed first: an end-to-end Core ML model run on Linux
The release announcement says Parakeet uses the Apple Neural Engine for its encoder and matches the macOS transcript byte for byte for on-device voice.
TLDR
The mlx-omarchy v0.5.0 announcement claims a Core ML model has run end to end on Linux for the first time. It says Parakeet’s encoder runs on the Apple Neural Engine, with its on-device voice transcript matching macOS byte for byte. The announcement lists M1 and M1 Max as fully supported, while M2+ support remains a work in progress.
Combined views
802
1 Source, first seen 14d ago