ModelHub

If you've ever tried to run local LLMs on a Mac, you know the pain. Your models are scattered across LM Studio, the HuggingFace cache, Ollama's blob storage, and a few random folders you forgot about. Finding, downloading, or removing one means jumping between browser tabs and terminal commands like it's 1995. Sabesh at Conscious Engines got tired of it and built ModelHub—the menu bar app that brings your entire LLM stack into one place.

ModelHub lives in your menu bar, plays nice with every backend—Ollama, MLX-LM, llama.cpp, transformers—and mirrors the official HuggingFace cache layout byte-for-byte. That means zero lock-in. Fetch a model through ModelHub and it's identical to one downloaded with huggingface-cli. Uninstall ModelHub tomorrow, keep your models. Version 1.4.0, barely 4MB, Apple Silicon only.

This is the app everyone who runs local LLMs didn't know they needed until they tried it. The problem it solves sounds small until you're staring at four different model directories wondering which one is the 3B Llama. Small tool, focused solves.