What is a model card primarily used for?

• A. A code repository containing deployment scripts.
• B. Documentation describing model purpose, limitations, and performance.
• C. A data labeling guideline document.
• D. A safety and compliance manual for users.

Answer: (B)

Model cards are concise documents that summarize what a model is intended to do, how it performs, and where its limitations lie. They provide essential context about the model—its purpose, the data it was trained on, the environments where it works well, potential risks or biases, and recommended usage constraints. This kind of transparent documentation helps teams make informed decisions about deployment, governance, and safety, and it supports accountability for regulators, stakeholders, and users. Key elements typically include the intended use, performance metrics and evaluation data, data provenance, potential limitations and biases, and deployment or maintenance considerations. By capturing these aspects in one place, a model card makes it easier to understand what the model can and cannot do, and where caution is needed.

The other options describe things that serve different purposes: a code repository with deployment scripts focuses on operationalizing the model, not on communicating its capabilities and risks; a data labeling guideline document guides how data should be labeled, not how the model behaves; and a safety and compliance manual for users covers broad safety and regulatory guidance rather than model-specific documentation.