Performance-based design optimization of steel frames, with element sections selected from standard sections, is a computationally intensive task. In this article, an efficient discrete optimization algorithm is proposed for performance-based design optimization of steel frames. The computational efficiency is improved by searching in a sensible manner, guided by the deformation information of structural elements. To include all standard sections in the design space, the cross-sectional area ( Area) and moment of inertia ( Ix) are selected as the design variables. Based on different relationships between Area and Ix, a twofold strategy is put forward, which includes a quick exploration and an elaborate exploitation. For comparison, a similar algorithm is also proposed, using Area as the only design variable. A fixed relationship between Area and other sectional properties is used. Two numerical examples are presented to minimize the structural weight while satisfying performance constraints. The results indicate that the proposed discrete algorithm can achieve lighter structural designs than the area-only algorithm. Furthermore, the convergence history proves that a high computational efficiency can be realized by using the proposed algorithm.