I realise that the difference lies in the way the defuzzification happens but I don't fully understand it. I've read some papers comparing the outputs from the two models but I'm still not really sure how they are different.
Mamdani type fuzzy inference gives an output that is a fuzzy set. Sugeno-type inference gives an output that is either constant or a linear (weighted) mathematical expression.
e.g Mamdani: If A is X1, and B is X2, then C is X3. (X1, X2, X3 are fuzzy sets).
Sugeno: If A is X1 and B is X2 then C = ax1 + bx2 + c (linear expression) (a,b,and c are constants)