Excessive withdrawal of groundwater resources has led to significant quantitative and qualitative issues in most aquifers in Iran. Population growth and limited surface water resources have intensified pressure on groundwater reserves to meet drinking, agricultural, and industrial demands. Therefore, monitoring and simulation of groundwater level and quality are essential for sustainable aquifer management. The Qazvin Plain, as one of the country's important agricultural regions with high dependence on groundwater resources, was selected as the study area. Artificial intelligence models offer an efficient approach for prediction and simulation due to their ability to learn complex nonlinear patterns without requiring complete understanding of aquifer physical processes. The objective of this research was to simulate groundwater quality and level in Qazvin Plain using three ensemble learning models (RF, XGBoost, and GBR) and one instance-based model (KNN). Comparison of model performance (Performance evaluation, scatter plots, time series, feature importance, Taylor diagrams, box plots and radar charts) revealed that ensemble learning models (RF, XGBoost, and GBR) with NSE and R² values of approximately 0.996 and precise identification of pivotal relationships among data, demonstrated superior performance in simulating groundwater quality and level compared to the KNN model.