A new computer-aided methodology for design and optimization of steel structures based on hybrid genetic algorithm with gradient learning of the best individual has been reported. The optimum design problem is formulated as the structural-parametric mathematical programming task with Boolean, integer and real design variables. In this way, cross sectional sizes of structural members, node coordinates as well as topology parameters can be considered as design variables. The system of constraints includes load-carrying capacity and stiffness conditions for structural members and entire steel construction according to building standards and regulations. Architectural, technological and other requirements can be integrated to constraint system as well. Determination of purpose function takes into account design specifications and ability to formulate the analytical expression as function of design variables. Hybrid genetic algorithm based on the parallel operations of genetic operators and update gradient method was used for solving the structural-parametric optimization task. Proposed technique was realized with elaborated software. Numerical example with new optimal design decision of plane two-hinged transverse frame with lattice structural members demonstrates the effectiveness of the proposed optimization methodology.