AI, Developer Tool
mlx-optiq
Run local LLMs on Apple Silicon with sensitivity-driven quantization and serving.
About mlx-optiq
OptiQ provides a full local LLM stack optimized specifically for Apple Silicon hardware. It quantizes open-weights language models using mixed precision so they fit into Mac unified memory, then serves them through a native server compatible with standard APIs. Users can inspect models, run interactive chat sessions in a local browser workbench, and deploy offline terminal coding agents without relying on external cloud infrastructure.
Who is mlx-optiq for?
Developers, AI researchers, and software engineers working on Mac hardware who previously had to choose between heavy cloud API subscriptions or slow, resource-intensive local runtimes. It serves developers looking to run, fine-tune, and inspect open-source AI models directly on their Mac terminal and local browser offline.
How does mlx-optiq work?
First, install the package via pip and compress an open-weight model using sensitivity-driven quantization. Next, launch the local server process to expose OpenAI and Anthropic compatible endpoints on your Mac. Finally, connect your browser UI, local coding agent, or text editor to run queries and execute code offline.