Surface Response Metodolgy: What model should i use for a 2 factor design

I made a SRM for a chemistry experiment. The factors analyzed were 2 an i had 5 responses.
For the model can i use a cubic model or should i use a cuadratic model due to the number or factors analyzed?
For some responses the best fit is a cuadratic and for some other is the cubic. Is correct adjust the responses to different models or i had to adjust the data to one model because all the responses were for the same experiment?
Thank you in advice