mlx-lm
Fewer than 5 people recommend it so far, so there's no written verdict. Here is what commenters said.
2recommend
1warn against
Dec 2, 2025first mention
Oct 8, 2026latest mention
Why people recommend it
“Get into a venv, and run:”
“mlx-lm / mlx-vlm — great work by others”
Caveats and dissent
“That too was broken in mlx-lm (it crashed)”
“I am not going to use mlx-lm or lmstudio anymore.”
Often compared with
Threads where it came up
- I ran Gemma 4 as a local model in Codex CLI 2 mentions
- Ask HN: Got a Mac for running agents? what's your setup? 1 mention
- Fixed three bugs that made Qwen3.5-122B a daily driver on Mac Studio 1 mention
- Show HN: RunAnwhere – Faster AI Inference on Apple Silicon 1 mention
- Ggml.ai joins Hugging Face to ensure the long-term progress of Local AI 1 mention
- Trying VLLM Ideas on Apple Silicon with MLX (WIP) 1 mention
Also written as mlx_lm, MLX-LM. Quotes are verbatim; the verdict and theme labels are written by a model from these comments, and every clause links to its sources.
All comments 8
“m2 24gb macbook mlx-lm & mlx-vlm own harness”
“I'd assume mlx-lm is worth checking too.”
“Get into a venv, and run:”
“That too was broken in mlx-lm (it crashed)”
“mlx-lm is a general-purpose array computation framework”
“I am not going to use mlx-lm or lmstudio anymore.”
“mlx-lm / mlx-vlm — great work by others”
“AI in app - mlx_lm”