Jun Kim joins Hugging Face to support oMLX and the MLX community

Hugging Face has hired Jun Kim, the creator and maintainer of oMLX, to provide dedicated support and funding for the project and the broader MLX community. This move aims to streamline the deployment of local AI on Apple Silicon by improving the transition of models from the transformers library to MLX implementations.

Impact on oMLX Development

Jun Kim will continue to lead oMLX, which remains licensed under Apache 2.0. The transition of oMLX from a side project to a fully funded and maintained project under Hugging Face is intended to provide greater stability and accelerate the pace of development, allowing for better guidance of contributors and long-term planning.

Strategic Goals for the MLX Ecosystem

Hugging Face intends for oMLX to serve as a testbed for new ideas while leveraging existing foundational dependencies such as mlx-lm and mlx-vlm. The overarching goal is to remove barriers for the community to run local AI in various forms by providing necessary tools and building blocks.

To support this ecosystem, Hugging Face is focusing on the following areas:

  • Streamlining Model Transitions: A primary focus is accelerating the process of moving from a transformers model definition to a reference MLX implementation. Because the transformers library serves as the industry reference for ML model definitions, streamlining this pipeline allows different engines to consume these implementations and focus on their own unique features.
  • Collaborative Upstreaming: Hugging Face aims to upstream work to relevant projects where appropriate, strengthening existing collaborations with teams involved in mlx-lm, mlx-vlm, and LMStudio.

Context on MLX and Hugging Face

MLX is Apple's framework specifically optimized for local AI on Apple Silicon. Hugging Face serves as the central Hub where users discover and contribute MLX models, supporting a healthy ecosystem of open, local AI tools.

Sources