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

Jun Kim, creator and maintainer of oMLX, has joined Hugging Face to keep building for the MLX community. Hugging Face frames MLX — Apple’s framework for local AI on Apple Silicon — as central to its local-AI push, and the Hub as where people already find and ship MLX models.

Side-project maintainership is how a lot of the local stack survives until it doesn’t. Moving oMLX from a nights-and-weekends job to a funded seat is the difference between a repo that stalls and one that can guide contributors for years. The project stays Apache 2.0 with Jun still leading it.

We’re glad this landed at Hugging Face rather than disappearing into a closed product team. Local AI only works if the tools that make models runnable on a Mac keep pace with the models themselves. Hugging Face’s stated end goal — unblock running local AI in any shape, and provide the building blocks — matches what we want for useful AI that doesn’t require a cloud round-trip for every inference.

The concrete focus they name matters: a quicker path from a transformers model definition to a reference MLX implementation that different engines can consume. Transformers has become the reference for model definitions; MLX still needs a cleaner on-ramp so each engine can specialize instead of reinventing export. That’s the kind of plumbing work that doesn’t trend, but it decides whether open local stacks stay competitive.

They also want oMLX as a testbed that leans on mlx-lm, mlx-vlm, and collaborators like LM Studio — with upstreaming where it makes sense. We’re watching that collaboration closely. Healthy ecosystems share primitives; unhealthy ones fork forever and leave users with half-broken DMGs.

One real-world hitch already showed up in the comments: people reported the oMLX GitHub repo looking deleted or auto-banned shortly after the announcement. That’s a reminder that maintainer transitions are operationally messy. Stability isn’t just funding — it’s repos, releases, and a clear URL when someone asks for a build.

Our take: this is useful AI infrastructure news. Funded maintainership for local Apple Silicon tooling is a net good. The risk if it goes wrong is capture — Hugging Face becoming the only serious MLX home — but the Apache license and stated upstreaming intent are the right hedges. I’m watching whether the transformers-to-MLX bridge actually ships and whether the community repos stay open and findable.

Hugging Face announced the hire in a September 22, 2026 blog post by Pedro Cuenca. MLX has been a community focal point since Apple’s 2023 release; Hugging Face positions itself as the Hub for MLX models.