About

This tool helps you figure out which open-source LLMs your computer can run locally, and roughly how fast, before you download anything. Interactive memory and speed estimates run in your browser. Optional custom GGUF lookups retrieve public Hugging Face metadata through this site, and community benchmarks are stored and aggregated on the server. No account is required.

Estimates are derived from device specs and model architectures, not live benchmarks. See the methodology for details and limitations.

The catalog maintains itself: a daily release radar scans the Hugging Face Hub for significant new open models and adds them automatically, and a weekly refresh keeps every model's params, quantizations, and download sizes current — no manual updates.

When a model is too large for your machine, we may link to cloud GPU providers that can run it; the build advisor also links to hardware to buy. Some of those are affiliate links — if you sign up or buy through them we may earn a small commission, at no cost to you. As an Amazon Associate, we earn from qualifying purchases. Affiliate commissions do not change which models or machines we show, or how we rank them.

Hardware purchase recommendations in the build advisor list machines that meet your requirements, cheapest first. The model directory is ordered by current Arena reference ratings, with unrated models afterward. The calculator offers sorts by fit, Arena reference rating, estimated speed and parameter count. See our privacy & terms for what we collect.