Local runtime - Apple (ml-explore) - verified
MLX-LM: install, API, platforms and features
MLX-LM is a local LLM runtime from Apple (ml-explore). As its docs describe it: A Python package for generating text and fine-tuning large language models on Apple silicon with MLX, offering a CLI, a Python API and a basic local HTTP server. The latest release is v0.32.2, published October 16, 2026. License: MIT (open source).
- Latest release
- v0.32.2
- Released
- 2026-10-16
- License
- Open source
- OpenAI API
- Yes
Install
Copied verbatim from the page linked under each command. Labels are worded as the source words them.
macos (Apple silicon, conda)
conda install -c conda-forge mlx-lmSource: github.com/ml-explore/mlx-lm
Run a model
The first-run example from the official docs. OpenAI-compatible endpoint: http://localhost:8080/v1 (mlx_lm.server listens on port 8080 of localhost; the HTTP API is described as similar to the OpenAI chat API and is not recommended for production)
mlx_lm.server --model mlx-community/Mistral-7B-Instruct-v0.3-4bitSource: github.com/ml-explore/mlx-lm/blob/main/mlx_lm/SERVER.md
Features
What the MLX-LM docs say it supports. Not stated means the docs do not say either way.
| Feature | Value | Source |
|---|---|---|
| Interface | CLI, Server, Python library | See sources below |
| Platforms | macOS, Linux | github.com/ml-explore/mlx-lm/blob/main/pyproject.toml |
| Backends | Metal, CUDA, CPU | github.com/ml-explore/mlx-lm/blob/main/pyproject.toml |
| Model formats | MLX | github.com/ml-explore/mlx-lm |
| OpenAI-compatible API | Yeshttp://localhost:8080/v1 (mlx_lm.server listens on port 8080 of localhost; the HTTP API is described as similar to the OpenAI chat API and is not recommended for production) | github.com/ml-explore/mlx-lm/blob/main/mlx_lm/SERVER.md |
| Anthropic-compatible API | Not stated | Docs do not say |
| Tool calling | Not stated | Docs do not say |
| Structured outputs | Not stated | Docs do not say |
| Vision (image input) | Not stated | Docs do not say |
| Multi-GPU | Not stated | Docs do not say |
| MCP client | Not stated | Docs do not say |
| License | MITOpen source | github.com/ml-explore/mlx-lm/blob/main/LICENSE |
| Latest release | v0.32.2October 16, 2026 | github.com/ml-explore/mlx-lm/releases/tag/v0.32.2 |
Models to run with MLX-LM
Browse the open-weight models you can run locally for download sizes, licenses and memory fit; check which formats MLX-LM loads in the table above.
Compare MLX-LM
FAQ
Does MLX-LM have an OpenAI-compatible API?
Yes. Endpoint, per its docs: http://localhost:8080/v1 (mlx_lm.server listens on port 8080 of localhost; the HTTP API is described as similar to the OpenAI chat API and is not recommended for production). Source: https://github.com/ml-explore/mlx-lm/blob/main/mlx_lm/SERVER.md
Does MLX-LM have an Anthropic-compatible API?
Its official docs, read on October 18, 2026, do not say.
Which platforms and GPUs does MLX-LM support?
Its docs list macOS and Linux, with these backends: Metal, CUDA and CPU.
How do I install MLX-LM?
On macos (Apple silicon, pip): "pip install mlx-lm", as shown at https://github.com/ml-explore/mlx-lm. This page lists 2 install commands: macos (Apple silicon, pip) and macos (Apple silicon, conda).
Is MLX-LM open source?
Yes. It is released under MIT; the license is at https://github.com/ml-explore/mlx-lm/blob/main/LICENSE.
What is the latest version of MLX-LM?
v0.32.2, published October 16, 2026 (https://github.com/ml-explore/mlx-lm/releases/tag/v0.32.2), as of October 18, 2026.
Which model formats does MLX-LM run?
MLX, per https://github.com/ml-explore/mlx-lm
Sources
Last verified from the pages below.
- GitHub README - github.com
- GitHub Releases - github.com
- HTTP Model Server (SERVER.md) - github.com
- pyproject.toml (platform classifiers and CUDA/CPU extras) - github.com
- MLX Build and Install - ml-explore.github.io
- MLX Custom Metal Kernels - ml-explore.github.io
Newsletter
Local AI, every week
One email per week. New model releases, runtime updates and what they mean for what you build.
Keep exploring
Open-weight models you can run
Download sizes, licenses and the run command for each model.
All local LLM runtimes
Every runtime in one feature matrix, with sources.
Picking a runtime for coding agents
Ollama, LM Studio, vLLM and llama.cpp for agent workloads.
Setup guides
Point your coding tools at the model you picked.