This paper addresses the optimal synthesis and design of multistage flash ( MSF) evaporator systems. A detailed nonlinear programming ( NLP) model based on a superstructure developed previously by Mussati et al. [ Ind. Eng. Chem. Res. 2003, 42, 4828- 4829] has been appropriately reformulated to include more alternative configurations for the process. The new superstructure includes the number of stages and the stream flow patterns ( distillate, feed and discharge brine, extraction points, and recycle flow patterns). Therefore, the superstructure simultaneously embeds an enormous number of alternatives, including, of course, the three commonly operating modes for the evaporator: MSF-BR, MSF-M, and MSF-OT systems and their combinations, which, to date, have not been analyzed systematically. In addition, a rigorous mathematical model by stage and component, which also includes the geometric design of each stage ( length, height, and width), number of tubes in the preheaters, fluid-dynamic equations for the streams among others, is applied. An attractive configuration for the MSF system resulted from the proposed superstructure. This structure differs from the conventional structure, because it considers distillate extractions, two recycle streams, and new allocations for the cooling seawater, blow-down brine, and recycle. The mathematical model and the solution procedure have been implemented in GAMS [ Brooke et al., GAMS: A User's Guide; Scientific Press: 1992]. Two study cases are presented, to illustrate the model and solution procedure capabilities. A complete description of the novel configuration, detailed comparison between different "sub-optimal" structures, and a sensitivity analysis on the main process variables are summarized.