A model card is the standing document that states what a model is for, what it was trained and tested on, how it performed and where it should not be used. It travels with the model through every version.
A card names the target, the filter, the training window, the evaluation period and the validation results, with enough detail that a model-risk reviewer can reproduce the claim. It records the known limits: forums, case families or input ranges where the model has little support.
The card is versioned with the model. When a model is retrained or demoted, the card is updated and the earlier card is kept, so a decision that relied on version two can be reviewed against the card for version two.
A model card is not marketing and not a certificate of quality. It is the disclosure a counterparty needs to decide how much weight an output deserves.
Working a file that turns on this term?
Working through a diligence process?
Institutional partners evaluating a position against this platform's outcome and duration models are welcome to reach out directly.