JANGQ AI's app, called MLX Studio on its site and installed as vMLX, runs open models with Apple's MLX. It has chat with vision models, built-in coding tools with MCP, image generation and editing, model conversion and a local API server. The vMLX engine underneath can also be installed on its own with pip.
It needs an Apple silicon Mac (M1 or later) on macOS 14.5 or newer, with 8 GB of memory at minimum and 16 GB or more recommended.
It's free, and the engine is open source (Apache 2.0). Inference runs on your Mac; the app only goes online to download models.