Simultaneous application of mechanical pressure and electrical charge on powder samples in spark plasma sintering process, has resulted in a sample with a density close to the theory. In the present study, a thermal-electrical-mechanical coupled finite element model of spark plasma sintering system using multi-objective optimization algorithm is proposed to optimize the mold variable. The simulation performed for Si3N4-SiO2 (1:1 mol) specimen has good agreement with the experimental results. Multi-objective genetic algorithms was used for optimization of mold design in order to maximize the temperature of sample core and minimize the mises stress in the mold. The results show that the optimized dimensions cause 8% increase in sample temperature and about 18% decrease in temperature difference between mold surface and sample core. This leads to better uniformity in the porosity distribution of final sample.