This study introduces the Exo-Spacetimeformer, an innovative adaptation of the Spacetimeformer model, designed to enhance PM2.5 air quality forecasting by incorporating external factors such as weather conditions, traffic intensity, and air-water-soil temperature differences into its decoder input. Through attention heatmap analysis and performance comparison across different time intervals, the Exo-Spacetimeformer demonstrates improved predictive accuracy over the original model, offering a promising tool for more reliable and comprehensive air pollution forecasting.