This paper describes a parametric study on window frame geometry with the goal of designing frames with very good thermal properties. Four different parametric frame models are introduced, described by a number of variables, such as dimensions of particular parts of the frame and thermal conductivity of the materials. In the first part of the study, a process of sensitivity analysis is conducted to determine which of the parameters describing the frame have the highest impact on its thermal performance. Afterwards, an optimization process is conducted on each frame. An attempt is made to optimize the design with regard to three objectives: minimizing the heat flow through the frame, maximizing the net energy gain factor and minimizing the material use. Since the objectives contradict each other, it was found that it is not possible to find a single solution that satisfies all of them. Instead, a range of semi-optimal solutions can be identified, from which the designer can select, according to their needs. A genetic algorithm was successfully used to address this problem. In the final part of the study, detailed simulations of energy use in a building are conducted to validate the results based on simplified, static simulations.