At the pinnacle of civilization, where the impacts of climate change have been increasingly felt, weather prediction plays a critical role in mitigating the potential disasters that may arise. Moreover, with the gradual change on climate, surface temperature of the earth is increasing. This increasing rate of the surface temperature causing global warming which is a matter of intimidation. To leave off this global warming, weather forecasting can be used as an arsenal. Selecting the appropriate tools and models for weather prediction is a crucial step in ensuring accurate forecasts. In this research paper, the focus was on studying the versatility of three specific architectures for weather prediction: LSTM, Temporal Fusion Transformer, and N-BEATS. To assess these architectures’ performance, we conducted a number of experiments. With the lowest Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) of the three, NBEATS stood out. This shows that when compared to the other models, the N-BEATS architecture had greater prediction accuracy. It's vital to remember, too, that the trials also showed that the Temporal Fusion Transformer and LSTM performed well. The only distinction was that these models required larger sizes in terms of parameters and computational complexity to achieve their performance levels. Consequently, considering both performance and model size, the researchers determined that N-BEATS was the most optimal and versatile architecture for weather prediction. Its ability to achieve excellent results with a smaller model size makes it a favorable choice for practical applications.