Model card

A model card is structured documentation describing a model’s intended uses, evaluation, limitations, development context, and relevant risks.

Plan for this page

Understand one concept in plain language and connect it to nearby ideas.

You will leave with: Useful cards identify version, model family, data or training summary where available, tested tasks and groups, metrics, ethical considerations, and out-of-scope uses. They record evidence supplied by the developer rather than independently certifying the model.
Time: 3–8 min
Before you begin: No prerequisite. Keep one real example in mind.
Do this now: Read the explanation, then compare the example with the common misunderstanding.

Prepared next stepLearn to inspect evidence status

Model cardA model card is structured documentation describing a model’s intended uses, evaluation, limitations, development context, and relevant risks.

Evaluation
Transparency
Provenance

Plain explanation

Useful cards identify version, model family, data or training summary where available, tested tasks and groups, metrics, ethical considerations, and out-of-scope uses. They record evidence supplied by the developer rather than independently certifying the model.

Example

A speech model’s card reports tested languages, error rates by setting, known weaknesses in noisy audio, license, and prohibited uses.

Why it matters

Documentation helps users compare claims with actual evidence and choose further tests. Missing, vague, or outdated information is itself important to note.

Common misunderstanding

Misconception: publishing a model card proves the model is safe and unbiased. Correction: the card is a transparency artifact whose claims and omissions still need review.

Related concepts

Follow relationships, not only the alphabet

EvaluationEvaluation is the structured measurement and investigation of how a model or system performs against defined questions, criteria, and contexts.TransparencyTransparency is making relevant information about a system, its use, evidence, limits, and governance available to the people who need it.ProvenanceProvenance is the recorded origin, custody, transformation, and history of data, media, evidence, or a system artifact.

Learn to inspect evidence status

Sources and scope

What supports this object

Artificial Intelligence Risk Management Framework (AI RMF 1.0)U.S. National Institute of Standards and Technology

Supports documentation, transparency, measurement and accountability across an AI lifecycle; document names and required contents vary by organization and jurisdiction.

Open source

Loading page content...

…