What it is
oMLX runs language, vision, embedding, OCR and reranker models on Apple Silicon through MLX and exposes them over an OpenAI-compatible API. It uses continuous batching and keeps KV cache in a memory tier and an SSD tier, so long contexts stay reusable across requests. A menu bar app and a web admin dashboard handle model downloads, pinning, auto-swapping, benchmarks and chat. Mac users running local models behind coding agents and other OpenAI clients use it as their always-on backend.