Strengths
- Requires minimal code to compare dozens of models (as few as three lines).
- Automatically handles both numerical and categorical features, including cardinality-based processing.
- Provides multiple performance metrics and training time for each model to aid comparison.
- Integrates with MLflow for experiment tracking.
- Free and open-source under the MIT license.
Limitations
- Uses basic models without parameter tuning, so results may not reflect optimized model performance.
- Last major updates were in 2021; some models like CatBoost have been removed.
- Limited to scikit-learn compatible models and does not support advanced custom architectures.