Mozi (@yeahhe)【教程】 Mac 端 LMStudio 本地部署 Qwen3.5-9B-MLX-4bit 中发帖

下载模型 
https://huggingface.co/mlx-community/Qwen3.5-9B-MLX-4bit

[PixPin_2026-03-03_02-29-24]
上下文拉满
[PixPin_2026-03-03_02-33-22]
关闭思考方法
顶部加一行
{%- set enable_thinking = false -%}

[PixPin_2026-03-03_02-23-09]
效果
[image]
[PixPin_2026-03-03_02-42-43]
开启 API 服务
lms server start --port 1234

Mac mini M4,功耗 40W,速度 21t 左右,多模态很强,内存占 5G 左右,普通聊天首字1s