This study investigates the use of machine learning (ML) techniques to predict sea surface temperature (SST) and marine heatwaves (MHWs) in 16 regions of the Mediterranean Sea. The ML algorithms, including random forest (RForest), long short-term memory (LSTM), and convolutional neural network (CNN), are trained and tested using datasets from the European Space Agency Climate Change Initiative (ESA CCI SST) and the European Centre for Medium-Range Weather Forecasts (ECMWF) ERA5. The models aim to forecast SST and MHWs up to 7 days ahead, with a focus on computational efficiency and accuracy. The results show that all ML methods performed well, with minimum root mean square errors (RMSEs) of about 0.1 °C at a 1-day lead time and maximum values of about 0.8 °C at a 7-day lead time. LSTM consistently outperformed CNN in most regions, and both LSTM and RForest outperformed CNN across all lead times. The ML techniques showed results similar to the dynamical Copernicus Mediterranean Forecasting System (MedFS) for both SST and MHW forecasts, especially in the early forecast days. For MHW forecasting, ML methods compared favorably with MedFS up to 3 days in 14 regions, while MedFS showed superior skill at 5 days in 9 out of 16 regions. All methods predicted MHW occurrence with a confidence level greater than 50% in each region. The study also highlights the importance of incoming solar radiation as a significant predictor of SST variability. The findings suggest that ML methods can be effective tools for short-term SST and MHW forecasting, providing valuable information for marine management and adaptation strategies.This study investigates the use of machine learning (ML) techniques to predict sea surface temperature (SST) and marine heatwaves (MHWs) in 16 regions of the Mediterranean Sea. The ML algorithms, including random forest (RForest), long short-term memory (LSTM), and convolutional neural network (CNN), are trained and tested using datasets from the European Space Agency Climate Change Initiative (ESA CCI SST) and the European Centre for Medium-Range Weather Forecasts (ECMWF) ERA5. The models aim to forecast SST and MHWs up to 7 days ahead, with a focus on computational efficiency and accuracy. The results show that all ML methods performed well, with minimum root mean square errors (RMSEs) of about 0.1 °C at a 1-day lead time and maximum values of about 0.8 °C at a 7-day lead time. LSTM consistently outperformed CNN in most regions, and both LSTM and RForest outperformed CNN across all lead times. The ML techniques showed results similar to the dynamical Copernicus Mediterranean Forecasting System (MedFS) for both SST and MHW forecasts, especially in the early forecast days. For MHW forecasting, ML methods compared favorably with MedFS up to 3 days in 14 regions, while MedFS showed superior skill at 5 days in 9 out of 16 regions. All methods predicted MHW occurrence with a confidence level greater than 50% in each region. The study also highlights the importance of incoming solar radiation as a significant predictor of SST variability. The findings suggest that ML methods can be effective tools for short-term SST and MHW forecasting, providing valuable information for marine management and adaptation strategies.