mlx-lm

Release: v0.32.2 (2026-10-16)Commit: 93e6d4aLicense: MIT

Run and fine-tune language models on Apple silicon with MLX.

inference

v0.32.2releasemlx-lmPythonMIT

FactsLicenseMITLanguagePythonKindinferenceRelease2026-10-16PlatformsmacOS (Apple silicon)Links to the projectpermissive

$ pip install mlx-lm
# use with Freeweights
$ mlx_lm.generate --model <id> --prompt "Hello"

Overview

Run and fine-tune language models on Apple silicon with MLX.

Topicsllmsmlx

Freeweights links to the project and does not host tool binaries.

$ pip install mlx-lm
# use with Freeweights
$ mlx_lm.generate --model <id> --prompt "Hello"

$ pip install mlx-lm

Checked against the project's README on 2026-11-14. The project's own instructions win when they differ: install docs

Freeweights links to the project and does not host tool binaries.

Run a catalog model

$ mlx_lm.generate --model <id> --prompt "Hello"

Download the weights from Freeweights first (HF_ENDPOINT or fw.py), then point the tool at the folder.

Freeweights links to the project and does not host tool binaries.

From:

ml-explore <github.com/ml-explore>Newsgroups: alt.freeweights.toolsSubject: mlx-lmRelease: v0.32.2 (2026-10-16)Commit: 93e6d4aLicense: MIT

Freeweights links to the project and does not host tool binaries.

Get it

Works with

1 catalog model. See the list