← back to the release
new featureData 360 AI Models
Predict Numeric Outcomes with Advanced Regression Loss Functions
Model Builder regression models can now use the Gamma or Quantile loss function instead of only squared error, so a model can be tuned for right-skewed targets (customer spend, claim amounts) or for a specific percentile (90th percentile of delivery time) rather than the average. Quantile models are evaluated with Pinball Loss and Coverage; Gamma and Squared Error models use the standard MAE, MSE, RMSE, and R2 metrics. Choose the loss function during model setup on the AI Models tab in Data Cloud.
identifiers
Model BuilderAI ModelsGamma lossQuantile loss
flags
enabled by defaultno action required
- how
- On the AI Models tab in Data Cloud, add a predictive model with regression capability, choose an algorithm that supports Gamma or Quantile loss, then select the loss function that matches your prediction goal.
- who
- when
- Available since August 2026.
- where
- Lightning Experience; Enterprise, Performance, Unlimited, and Developer editions.