The 2026.07 release of SAS Model Manager includes some BIG changes for the model card. The changes include expanding the scope of the model card, adding the ability to export the model card outside of SAS Viya, and making it easier to generate the assessment metrics for a full model card. There will be an article deep-diving into each new enhancement, with this article focusing on our new lighter version of the model card that will be available for any and all models stored in SAS Model Manager.
Before the 2026.07 release, only classification and prediction models with a full set of assessment metrics generated through SAS Model Studio, SAS Data and AI Studio (formerly called SAS Studio), or python-sasctl, had a model card in SAS Model Manager. These metrics populated the donut charts on the Overview tab and the Model Summary section. One of the changes in the 2026.07 release was to create a lighter version of the model card that could be displayed when these metrics weren’t available. But don’t worry -- if these metrics are available, the full model card experience will still be shown.
This lighter model card will still include an Overview, a Data Summary, and a Model Usage section so teams can document the model purpose, limitations, and out-of-scope use cases as well as review the training data.
For teams that want to add additional information to their model card, all model cards now support the ability to add custom charts from SAS Visual Analytics. To add custom visualizations to your model card, open the Properties tab for the model, then select the new Model Card Settings section. In this section, you can specify to show the Custom Charts section of the model card and select Add to pick the report from SAS Visual Analytics to use. After saving and refreshing your model, there will be a Custom Charts section of the model card that displays your report from SAS Visual Analytics.
Other additional enhancements include the ability to document the model’s lifecycle status as it moves through validation, deployment, retraining, and retirement as well as the model’s approval status between awaiting approval, approved, rejected, and retired. A similar selection mechanism for adding the report from SAS Visual Analytics to the model card was also adopted to select and add a custom chart to the performance monitoring report, reducing the number of steps it takes to add a custom chart.
This demo explores the new lighter model card as well as reviews how to customize all model cards using SAS Visual Analytics:
The next article will focus on exporting the model card, so stay tuned!
