Compare model quality without leaving the clinical workspace.

The model lab reads the live comparison endpoint and keeps the metric story explicit. If data is missing, the interface says so directly instead of inventing confidence.