Ever found a shiny new model on Hugging Face, got super excited, downloaded it, and then watched your GPU cry as it immediately ran out of memory? Yeah, me too.
That's why I built this.
Hugging Models is a lightweight, native macOS menu bar app that lets you search Hugging Face, explore trending models, and instantly calculate the estimated VRAM required to run them before you nuke your system.
- 🔍 Direct Hub Search: Search Hugging Face models right from your menu bar.
- ⚡️ Trending Indicators: See what's hot right now. We pull the exact same momentum signals Hugging Face uses to highlight trending models.
- 📊 Metrics at a Glance: Instantly view downloads and community likes for any model.
- 🧠 Smart Auto-Detection: Automatically pulls architectural configs (layers, heads, dimensions) and guesses the parameter count if it's in the model name.
- 📦 GGUF Support: Detects GGUF variants and allows you to select specific quantization files to see precise memory requirements.
- 🧮 KV Cache Math: It doesn't just calculate weight size; it estimates the KV cache overhead based on your desired context length and batch size.
If you want to compile this yourself, you'll need a Mac and Swift (Xcode command line tools are fine).
- Clone the repo.
- Open your terminal and navigate to the
HuggingModelsdirectory. - Run the magical
makecommand:
make all- Find your shiny new
HuggingModels.appinside thebuild/folder. - (Optional) Run
make runto build and immediately launch the app. - (Optional) Run
make installto copy the app to your/Applicationsfolder.
It's a clean, standard SwiftUI MVVM architecture. If you want to poke around, start at ViewModel.swift (the brains) or ContentView.swift (the beauty). Check out ARCHITECTURE.md for the fancy Mermaid diagrams.
This project is licensed under the Apache License 2.0. Do whatever you want with it, just don't blame me if your GPU still catches fire.