# ml-explore/mlx-lm

Run LLMs with MLX

Repository: https://github.com/ml-explore/mlx-lm
Canonical: https://ross.abutalabs.com/products/mlx-lm
Language: Python
License: MIT
License Family: permissive
Topics: llms, mlx
Last push: 2026-08-26T22:58:05+00:00

## Health v2 (maintenance only)
Score: 80/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 80, longevity 38
- inputs: {"age_days": 540, "days_push": 7, "days_rel": 133, "gap_med": 10, "n_releases_24m": 28}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 6801, forks 996 (observed 2026-08-28T04:09:48.653301+00:00)

## What it is
A Python package for running and fine-tuning large language models on Apple silicon using the MLX framework. It integrates with the Hugging Face Hub for model loading, quantization, and uploading, and provides both CLI tools and a Python API for text generation, chat, and LoRA/full fine-tuning.

## Use cases
- run llms locally on mac
- generate text with a local llm on apple silicon
- fine-tune an llm with lora on a mac
- quantize a hugging face model to 4-bit
- chat with a local language model in the terminal
- convert and upload models to the hugging face hub
- run distributed llm inference on apple silicon

## When to choose
- you want to run or fine-tune LLMs on Apple silicon (M-series) hardware
- you want easy access to thousands of Hugging Face models with quantization support
- you need both CLI tools and a Python API for LLM inference

## When to avoid
- you are on Linux, Windows, or CUDA GPUs - use llama.cpp, vLLM, or transformers instead
- you need production-grade multi-GPU serving rather than local Apple silicon inference

## Facets
- artifact type: library
- maturity: active
- function: llm-inference, llm-training, cli, sdk
- domain: large-language-models, machine-learning, developer-tools
- platform: python, cli
- tags: apple-silicon, mlx, hugging-face, fine-tuning, lora, quantization, text-generation, macos

## Member repositories
- ml-explore/mlx-lm (main) score 80

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:09:48.653301+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:42:05.458603+00:00, confidence not recorded.
  - readme: https://github.com/ml-explore/mlx-lm (fetched 2026-08-28T04:09:48.653301+00:00, sha c6731cb2bc49)
  - registry_pypi: https://pypi.org/pypi/mlx-lm/json (fetched 2026-08-29T08:38:07.797887+00:00, sha 38dfb91a39ec)
- Data as of 2026-08-30T08:39:29.467469+00:00.