AI Rank

By Wednesday, three newsletters have already called Monday's new AI model the best one yet. Loud launches make it hard to see which projects people keep using once the noise dies down. AI Rank builds its daily lists from usage figures instead. Most of them come from GitHub, where the bulk of open-source code lives, and from OpenRouter, a service that sells access to many AI models through one account.

The open-source board orders code projects by the stars they gained since the previous day's snapshot and picks out the fastest risers. The site says it follows 22,535 AI repositories.

Models are ranked by how much text they read and write on OpenRouter, measured in tokens. A second model list follows blind votes on Arena, where people judge two anonymous answers side by side. Another board sorts coding agents and other AI apps by token use and request counts.

Each day a short brief sums up what moved, next to sections for AI news and longer articles, and the whole site also comes in Chinese. The iPhone app adds rank history for each model and comparisons between providers, and extra filters come with an optional subscription. The usage ranking of models draws on OpenRouter traffic alone, so it reflects only what runs through that one service.