In order to optimize the performance of modified high-performance building gypsum, a 0.20% vegetable protein retarder was used to delay the setting time of the building gypsum. The response surface methodology was employed to design a series of experiments involving the combined use of three admixtures: polyether modified silicone defoamer, redispersible latex powder, and hydroxypropyl methylcellulose. The experimental response indicators were analyzed, and the Design-Expert 12 software was used to set optimization conditions for these response indicators. Based on the Optimization function, optimal parameters for the dosage of each admixture were generated, and the feasibility of the optimization results was verified. The results indicated that the optimal admixture dosages for the high-performance building gypsum were 1.051% polyether modified silicone defoamer, 0.827% redispersible latex powder, and 0.141% hydroxypropyl methylcellulose. When comparing the modified high-performance building gypsum with the unmodified version, significant improvements were observed in six performance metrics: absolute dry flexural strength, absolute dry compressive strength, tensile bond strength, water retention rate, softening coefficient, and porosity.