A performance‐based framework is presented to investigate the optimum tuning of tuned mass dampers (TMDs) for seismic applications in several steel moment‐resisting frames by single and multi‐objective genetic algorithm. The single objective problems have been developed for a set of given mass ratios so as to meet the best performance, while the multi‐objective problems try to attain a set of Pareto optimal designs so as to find the relation between TMD efficiency and TMD mass. In the proposed approach, nonlinear time history analysis via endurance time (ET) method is utilized to estimate the structural responses at two seismic hazard levels in order to assess the performance of structures, while a new criterion for scaling period of TMD‐equipped structures is proposed to be used in scaling process of seismic excitations. A criterion of interstory drift ratios at BSE‐1N seismic hazard level, as the performance index of the studied frames, is considered to be minimized and finally, detuning effects of the optimum TMD designs are checked at BSE‐2N seismic hazard level. In order to illustrate the superiority of the proposed approach, the robustness of optimal TMD parameters is evaluated under near‐ and far‐field ground motion record sets recommended in FEMA‐P695. Also, the optimum results are compared to those from several analytical methods. The proposed framework, taking the advantage of the ET method, is found to be an efficient and feasible approach for practical and cost‐effective optimum seismic design of TMDs on nonlinear structures with a reasonable accuracy and significantly reduced computational effort.