🔍 Read the full analysis: Jun Kim Brings oMLX Experience To Hugging Face’s MLX Community on ThorstenMeyerAI.com
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Jun Kim, creator and maintainer of the open-source oMLX inference engine, has joined Hugging Face to work on Apple’s MLX ecosystem. Hugging Face says oMLX will receive funded maintenance while staying Apache 2.0 licensed and under Kim’s leadership; the company has not given a timeline for the technical work it described.
Jun Kim, creator and maintainer of oMLX, has joined Hugging Face to work on Apple’s MLX ecosystem, the company said in a blog post. Hugging Face says the open-source inference engine will move from a side project to a funded, fully maintained effort, while remaining under Kim’s leadership and Apache 2.0 license.
Hugging Face said the new support is intended to give oMLX more stability and faster development. The company said funding the work will help Kim guide contributors and maintain the project over the long term. The post did not specify the amount or form of funding, or whether it will cover particular development milestones.
The company also described a technical priority: making it simpler to turn a model defined in the Transformers library into a reference implementation for MLX that multiple inference engines can use. Hugging Face said this could make it easier to run newer Transformers models on MLX. It did not publish a schedule, name specific models, or set out a delivery plan for that work.
Hugging Face said Kim will continue to lead oMLX, and that the project’s Apache 2.0 license will remain. The company described oMLX as building on libraries including mlx-lm and mlx-vlm, and said it hopes to collaborate with their teams and with LMStudio. It also said it would be open to upstreaming suitable work into other parts of the MLX ecosystem.
More Support for Apple Silicon Inference
For developers who run models locally on Apple hardware, a funded maintainer could make an established community tool less dependent on spare time. Single-maintainer projects can face interruptions when their maintainers have limited time; Hugging Face’s announcement addresses that risk by placing Kim’s oMLX work in a funded role. Whether that support results in faster releases or improved reliability remains to be seen.
The move also connects oMLX to a company that operates a major model-sharing platform and maintains the Transformers library. That connection could help developers move between model definitions and MLX implementations. Hugging Face has described this as a goal, though it has not announced completed integrations or specific changes users can access today.
The project’s license and leadership terms are relevant to contributors who want to understand how its governance may change. Hugging Face says both will continue as before. Its interest in upstreaming work and working with other MLX teams could also benefit projects beyond oMLX if collaboration produces shared code. For now, that broader effect is a stated intention rather than a confirmed result.
How oMLX Fits Into the MLX Ecosystem
MLX is Apple’s machine-learning framework designed for Apple Silicon. Hugging Face’s post dates its support for MLX to the framework’s release in late 2023, which the company called “the Christmas present from Awni and Angelos in 2023.” MLX-related projects such as mlx-lm and mlx-vlm provide tools and model implementations on top of the framework.
oMLX is an inference engine in this ecosystem, created and maintained by Kim. Hugging Face’s Hub is a place where users can find and publish MLX-format models. The company said it has collaborated with MLX-related projects, including mlx-lm, mlx-vlm and LMStudio, and wants to strengthen ties with the teams behind them. Those details place Kim’s hire within Hugging Face’s existing work around local AI on Apple hardware, rather than describing a new MLX framework or a change to the Hub.
Hugging Face characterized local and open AI use as accelerating and said MLX is central to that ecosystem on Apple devices. Those are the company’s assessments; the announcement did not include usage figures or a comparison period. Its concrete commitments in the post concern Kim’s role, continued project leadership and licensing, alongside a stated focus on making Transformers models easier to bring to MLX.
““We are completely invested in local AI, and MLX is a central piece of the ecosystem.””
— Hugging Face, in its announcement
Details of Kim’s Role Remain Unspecified
The announcement does not give Kim’s job title, team placement or start date. It also does not describe funding terms, the resources assigned to oMLX, or whether the support has a fixed duration. The company has not published a schedule for the Transformers-to-MLX work or identified the first model implementations it expects to support.
Hugging Face’s comments about possible upstream contributions and collaboration with other MLX teams describe intentions, not specific commitments with named deliverables. It is not yet clear how those efforts will be organized, which changes will land in oMLX or other libraries, or when users might see them. The announcement does not say whether the project’s development priorities or release process will change.
The clearest indicators will be updates to the oMLX repository and release notes, which can show whether the promised funded maintenance leads to new releases or reliability work. Developers can also watch for documentation or code demonstrating a path from Transformers model definitions to MLX implementations usable by more than one inference engine.
Further updates may clarify Kim’s role, how Hugging Face will support the project, and whether work is shared with mlx-lm, mlx-vlm, LMStudio or other MLX teams. Hugging Face said it will continue supporting MLX models on the Hub, but gave no timeline for these efforts. Until concrete releases or collaboration details appear, the speed and reach of the changes remain open.
Key Questions
What did Hugging Face announce?
Jun Kim has joined Hugging Face to work on Apple’s MLX ecosystem and continue leading oMLX, the open-source inference engine he created.
Will oMLX change its license?
Hugging Face said oMLX will remain Apache 2.0 licensed. The announcement did not describe any planned licensing change.
What technical work is planned?
Hugging Face said it wants to make it easier to turn Transformers model definitions into reference MLX implementations that multiple inference engines can use. It did not give a timeline or name initial models.
When will users see changes to oMLX?
The company has not announced a release schedule or dates for the planned technical work. Repository updates and release notes will show what changes reach users.
Primary source: Hugging Face · via ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
