Methodology
Every model on this site is scored from public data. No vibes, no expert reviews, no pay-to-play.
Data sources
OpenRouter API — model catalog, pricing, context, modality, supported parameters. Refreshed daily.
Public model documentation — vendor spec sheets for things OpenRouter doesn't expose.
Composite quality score
A 0-100 score combining context window, pricing, modality breadth, tool-calling support, structured output support, and reasoning capability. Higher = stronger general-purpose model.
Per-task ranking
Each task page reweights the model dimensions for what that task actually needs. A "Best for SQL" page weights structured output and reasoning highly; a "Cheap bulk inference" page weights cost above all else. Top-15 per task are surfaced.
Affiliate disclosure
"Try via OpenRouter" links pass an affiliate parameter. If you sign up and spend on credits, OpenRouter shares a small fraction with this site. Rankings are not influenced by affiliate revenue — every model gets the same affiliate treatment regardless of its score.