Project AI

MLX

MLX, Apple's ML framework optimised for Apple Silicon. Underpins on-device inference on Mac and iPhone.

29k Stars

2.4k Forks

Releases tracked

GitHub releases

Source type

99 days

Avg cadence

Jul 2026

First tracked

7d ago

Latest

Release history

Adds VJP support for scaled dot product attention, needed for training/backprop workflows using MLX attention layers

Minor bump to 0.32.0; Qmm implementations now generated via cmake