Data sources

Where OpenModelStats' information comes from, and how source data differs from our own analytics.

Hugging Face Hub API

Initial model metadata — names, owners, task labels, download and like counters, license tags and related repository facts — originates from the official Hugging Face Hub API. Requests are authenticated, throttled and logged. OpenModelStats is an independent analytics platform and is not affiliated with or endorsed by Hugging Face.

Source metadata vs. OpenModelStats observations

  • Source metadata is whatever the platform reports right now: current counters, tags and timestamps. It can change at any time without notice.
  • OpenModelStats observations are the values we record during our own daily collection runs, together with everything derived from them: interval gains, velocity, rank movement and charts. These are produced solely by OpenModelStats.

Every page clearly separates these two categories. Wherever uncertainty exists — missing baselines, counter resets, partial coverage — we display nothing rather than guess.

Linking to the original platform

Every model page links directly to its original repository. OpenModelStats does not host model weights, mirror model cards, or republish repository assets.

Future sources

The architecture is designed to admit additional public model platforms over time. Today, Hugging Face is the sole active source; any future additions would be documented here before their data appears on the site.

Questions about the data? See the methodology or contact us.