Running Qwen3.6 and Qwen3.8 in LM Studio on an M5 Pro

Using 4-bit MLX models locally on Apple Silicon while managing unified memory, context length, and thinking settings.