MLX 0.32.4 Brings Targeted Attention Fixes to Older Apple Silicon
MLX 0.32.4 addresses an M1/M2 attention error, specific memory use on older chips, and a fast-sync hang. Check your runner before retesting.
MLX 0.32.4 addresses an M1/M2 attention error, specific memory use on older chips, and a fast-sync hang. Check your runner before retesting.
Liquid AI’s LFM2.5-24B-A2B targets local agents on Apple silicon. Its MLX version requires mlx-lm 0.32.0 or later to route experts correctly.