MLX Harmony: Notes from an Earlier Experiment
MLX Harmony was an experiment I used to explore GPT-OSS on MLX and the Harmony prompting structure. It was useful as a place to learn how the engine and prompt format behaved together, but it did not become a model-hosting product.
What It Was
The repository collected a lightweight interface and the code needed to exercise that particular combination of model, MLX engine, and prompting structure. I am keeping the material available as an experimental record rather than presenting it as an active roadmap.
Where the Work Moved
My current attention is on MLX Forge. That project is concerned with model conversion, inference performance, and repeatable evidence about the changes that help on Apple Silicon. It remains private while I work through its initial milestones, so I am not treating this older experiment as a substitute project page.
Join the Discussion
Comments for this post live in GitHub Discussions. That keeps moderation in one place and gives the conversation a stable home.