Mac M1, llama.cpp and pi harness.
Mac M3, oMLX, pi. 64GB RAM. Qwen 3.8 27b / Gemma 12b
m2 24gb macbook mlx-lm & mlx-vlm own harness
Parakeet TDT 0.6B v3 (int8 CoreML, about 470 MB) runs on-device through FluidAudio's AsrManager, pinned to CPU plus the Neural Engine.