Hugging Face has hired oMLX creator Jun Kim, turning maintenance of the local-model server from a side project into funded full-time work. The company says oMLX will remain under the Apache 2.0 license, with Kim continuing to lead the project and its contributors.
oMLX belongs to the ecosystem around MLX, Apple’s machine-learning framework optimized for Apple silicon. It works with components such as mlx-lm and mlx-vlm to serve language and vision-language models locally on Macs, where memory and accelerator support differ from conventional GPU servers.
One planned focus is shortening the delay between a new model definition in Hugging Face Transformers and a reference MLX implementation. A common conversion path could let inference engines adopt newly released architectures without each project independently recreating all model logic. Hugging Face also says oMLX can act as a testbed whose useful changes may be contributed upstream.
The hire gives the project more stable maintenance but does not transfer it to a closed product or guarantee compatibility with every model. Progress will still depend on coordination with MLX libraries, engine developers and Apple’s hardware-specific tooling.