Jun Kim, the creator and maintainer of oMLX, has joined Hugging Face to work full time on the project. This move shifts the codebase from a side initiative to a fully funded effort, aiming to improve stability and accelerate development. Kim will continue to lead the project under the Apache 2.0 licence while guiding a growing team of contributors. The goal is to build a sustainable infrastructure for local artificial intelligence that runs efficiently on Apple Silicon devices.
This change matters because it addresses the resource constraints often faced by open source frameworks. By securing dedicated funding, the project can focus on long term planning rather than immediate maintenance. A primary objective is to create a direct bridge between the transformers library and MLX implementations. This allows developers to port existing models without rewriting code from scratch, reducing the barrier to entry for local inference. The strategy relies on upstreaming improvements to the wider ecosystem to benefit all users.
- oMLX remains Apache 2.0 licensed
- Focus is on transitioning transformers definitions to MLX
- Collaboration continues with teams behind mlx-lm and LMStudio
