BuildLearnCommunity ModelsCheapestLong-contextOpen sourceImage modelsTrending How we score AI tools API key
🎯 Don't want to pick? Send model: "appmo/auto" and Appmo routes each request to the cheapest model that's good enough — cutting your bill up to 40%. See how →
Models › How we score

How we score AI models

Every number in this directory, and where it comes from.
SignalWhat it measuresStatus
Price & usage rankLive per-token prices and weekly usage ranking from OpenRouter.weekly
Open weights & licenseA model counts as open weights when its weights are published on Hugging Face; the license comes from the model card.weekly
Downloads & likesHugging Face downloads (last 30 days) and likes — how many people actually run it themselves.weekly
Hacker NewsStories naming the model, with points and comments.weekly
GitHubNew public repos mentioning the model — a signal of real developer adoption.weekly
Buzz (0-100)Combined attention on Hacker News and GitHub, relative to the most-discussed model we track.updated 2026-11-08
Community sentimentA simple positive/negative word count over post titles and text, weighted by upvotes. Shown only with at least 5 relevant posts.weekly

Limits worth knowing

Community buzz measures attention, not quality — a model can be loudly discussed because it disappointed people. Sentiment is a word-count estimate, not a reading of every comment. Usage rank and price say what developers pay for, not how good an answer is. Image models are not yet scored on image quality.