Model card (a model's fact sheet)
Japanese モデルカード
A model card is a short document published with an AI model that explains what it is meant for, what data it learned from, how it scored in tests and where it falls short. It is the first thing to read before using a model.
A model card is a short fact sheet published alongside an AI model. It says what the model is meant to be used for, what data it was trained on, how it scored in evaluations, and what limits and biases it has. Think of it like the instruction manual that comes with a home appliance. On Hugging Face, the README file in a model's repository works as its model card, with details such as the license, languages and training datasets listed at the top.
Watch out The people who built the model write the card themselves, so strengths tend to come first. If a news story or blog post gives numbers that differ from the model card, check the card first. Before using a model for work, read the license field and the section on limitations.
Source: Model Cards · Model Cards for Model Reporting (checked 2026-10-26)