Native Swift + MLX port of FLUX.2 [klein] ships for Apple Silicon
A dependency-free Swift + MLX port of FLUX.2 [klein] for Apple Silicon adds text/image editing, quantization, and Python-parity outputs.
A developer has released a fully native Swift + MLX implementation of FLUX.2 [klein] that runs entirely on Apple Silicon with no Python, no diffusers, and no external dependencies. The project supports text-to-image, image-to-image, and mask-guided editing — including background replacement, object removal/addition, recoloring, and other semantic edits.
On the memory side, it ships with bf16/fp16/int8/int4 quantization, staged model residency, VAE tiling for high resolutions, and a low-memory mode with reporting. Notably, it achieves seed-42 parity with the official MLX Python reference, meaning outputs match exactly.
The repo includes a dependency-free CLI, a SwiftPM library, a minimal SwiftUI demo, docs, tests, and a contributor roadmap. The goal is a genuinely native diffusion pipeline that can be dropped into macOS/iOS apps without pulling in the Python ecosystem — a notable step for developers building local image generation and editing tools with MLX.