Model card
A short spec sheet for a model — what it’s for, what it was trained on, and where it falls down.
Think of it like
The nutrition label on food: not the meal, but what you should know before consuming it.
Example
Before adopting a model, a team reads its model card for the license, known limitations, and eval scores.
How it actually works
A model card documents intended use, training data at a high level, evaluation results, and known risks or biases. It’s a transparency and governance artifact more than a technical one. Quality varies wildly, but it’s the first thing to check before you build on something.
For product teams
Your first due-diligence read — license, intended use, and stated limits can make or break adoption.
For engineers
Structured documentation of a model’s scope, data, metrics, and limitations, per the model-cards convention.
Related
- Base model — Describes a base model.
- Evals — May cite evals.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome