The latest release of the GitHub repository ml-explore/mlx-swift-examples is 2.29.1. It is a repository that gathers example programs using MLX Swift, containing a chat app and an MNIST trainer that run on iOS and macOS, plus command-line tools for LLMs, image generation, and MNIST training.
The center of it is chat examples for LLMs and VLMs, and examples for viewing model downloads and inference statistics. A VLM means a model that takes images as input. The numerical computation examples are CurveFit, HeatTransfer, and Mandelbrot, while StableDiffusion and MLXMNIST are grouped separately as reusable libraries that can be imported and used outside the examples. The language is Swift and the license is MIT, and the repository, created on February 22, 2024, has 2684 stars.
That said, this repository is not a library but a collection of examples, and this record contains no changelog for the 2.29.1 release, so what changed cannot be known. There are no performance comparison figures in this record either; in the research record only the momentum score is filled in, at 1.0 out of 10, with the other entries blank, so judging recent spread on this basis is still premature. It is immediately useful as reference material for choosing examples, but too narrow to take a finished library and use it as-is.