shoehorn
Quantize any model down to run on your machine
Last verified:
What is shoehorn?
shoehorn is an open-source tool that quantizes and fits large language models into the memory you actually have, matching Hugging Face models to your hardware budget. Features a browser-based discovery interface and local web UI that ranks models by quality for your specific machine specifications and memory constraints.
shoehorn pricing
Pricing model: Freemium
Free (open-source)
shoehorn pros
- Hardware-aware model discovery ranked by optimal fit to your budget
- Privacy-first local inference with llama.cpp backend—no cloud dependency
- One-button model download and chat setup
- Multi-platform support: macOS (Apple Silicon), Linux (x86-64), Windows (x86-64)
shoehorn cons
- Requires llama.cpp installed on system PATH as a dependency
- Limited to Hugging Face's most-downloaded models for discovery
- Minimal documentation visible on this landing page
Frequently asked questions about shoehorn
How do I install shoehorn?
Via Homebrew (`brew install notactuallytreyanastasio/shoehorn/shoehorn`), direct binary download for your OS, or from source with `cargo install`.
What models can I run?
Hugging Face's most-downloaded models, ranked and filtered by what fits your available RAM/VRAM.
Does it require internet after setup?
Models are downloaded locally for offline inference, though the discovery browser interface scans Hugging Face initially.