Hugging Face has hired Jun Kim, the creator and maintainer of oMLX, to support the MLX community. The company announced the move on its blog on September 22, 2026.
MLX is Apple's framework for local AI, optimized for Apple Silicon. Hugging Face has backed it since its December 2023 release, running the main hub where developers find and publish MLX models. The company says it is "completely invested in local AI" and treats MLX as a central piece of that work.
What's new
- Jun Kim joins Hugging Face, taking oMLX from a side project to a fully maintained and funded effort — still Apache 2.0, still led by Kim.
- Hugging Face wants closer collaboration with existing MLX projects including mlx-lm, mlx-vlm and LMStudio, plus a smoother path from Hugging Face Transformers model definitions to reference MLX implementations that inference engines can consume.
- The company says the change should bring stability and faster development, letting Kim better guide contributors, and it is open to upstreaming work into dependencies where that makes sense.
Why it matters
Local AI on Apple Silicon has grown fast since MLX arrived, but the ecosystem has leaned on community-maintained bridges between Hugging Face's Transformers library and MLX-native implementations. By hiring the creator of one of the key translation layers, Hugging Face is investing directly in cutting friction for developers who want to run the latest Transformer models locally on Mac hardware. If it works, new architectures should reach MLX sooner.
Our take
Hugging Face is funding the maintainer of a critical community bridge rather than building a competing one. That is pragmatic: it validates the oMLX approach and gives it resources to keep pace with Transformers' release cadence. The test is whether upstreaming into mlx-lm and mlx-vlm actually happens — those projects have their own roadmaps and maintainers. Deep collaboration means fewer fragmented translation layers for developers; otherwise oMLX stays a well-funded but separate path.
