Model card
A standard document describing a model: intended use, training data, evaluation results, limitations and ethical considerations.
Why it matters. Tells you whether your use case is in scope before you build on the model.
How it works
A model card is a document describing a model: who made it, intended and out-of-scope uses, training data at a high level, evaluation results, known limitations and ethical considerations. The format was proposed in the research paper 'Model Cards for Model Reporting' (Mitchell et al.) and is now widely used.
The level of detail varies widely, from a short page to a long technical report. Open-weights releases often ship a card next to the weights, and some makers call their longer release report a system card instead.
Read the intended and out-of-scope uses before building, and check the limitations against your use case. On this site, card facts come from the maker's own card with page or section references.
Worked example
Illustrative: a card lists medical diagnosis as out of scope and notes weaker results in low-resource languages. A team planning a triage assistant in Swahili now knows to run its own evaluation, add clinician review, or choose another model.
Common mistakes
- Skipping the out-of-scope section and building exactly what it warns against.
- Assuming the card's evaluation results reflect your data.
- Not re-reading the card when a new model version ships.
Across the catalog today
Longest system and model cards we've summarized
- Claude Fable 5System Card: Claude Fable 5 & Claude Mythos 5317 pages
- Claude Opus 4.8System Card: Claude Opus 4.8246 pages
- Claude Opus 4.7System Card: Claude Opus 4.7232 pages
- Claude Opus 5.5System Card: Claude Opus 5.5230 pages
- Claude Opus 4.6System Card: Claude Opus 4.6213 pages
- Claude Fable 5.1System Card: Claude Fable 5.1 & Claude Mythos 5.1212 pages
See it on a real card: Overview on Claude Opus 5.5.