Hugging Face announced on September 22 that oMLX creator and maintainer Jun Kim has joined the company. Kim will continue leading the open-source project, which is part of the local AI ecosystem built around MLX, Apple’s framework optimized for Apple Silicon.
The company said the move will turn oMLX from a side project into a funded project with ongoing maintenance. Hugging Face expects the additional support to improve stability, potentially accelerate development and give Kim more capacity to coordinate contributors and plan the project’s long-term direction.
The project will remain under the Apache 2.0 license, and Kim will retain its leadership. The announcement therefore changes the resources and employment structure behind oMLX rather than moving it to a proprietary license or assigning its direction to a new maintainer.
Hugging Face plans to use oMLX as a testbed for new ideas while continuing to draw on foundational dependencies such as mlx-lm and mlx-vlm. It also wants to contribute suitable work back to related projects and strengthen collaboration with teams building other tools across the MLX ecosystem.
One stated focus is creating a faster path from a model definition in the Transformers library to a reference MLX implementation that multiple inference engines can consume. Hugging Face’s aim is to provide engines with a shared modeling foundation, allowing their developers to concentrate on distinct features while making new Transformers models easier to run through MLX.
Practical context: For developers, this is an organizational commitment rather than a new oMLX release: Hugging Face did not announce a version, delivery schedule or near-term feature list. A common route from Transformers definitions to reusable MLX implementations could reduce duplicated model-integration work across engines, but its practical impact will depend on future code changes and adoption by other projects.
Sources
Event date: 2026-09-22. Primary source date: 2026-09-22.
