This paper aims at solving the contradictive design problem of the modular steel frame structure in cold northern regions when considering both energy- and cost- saving. A synchronous optimization study with energy consumption and cost objectives is hence carried out for target modular steel frame structures. First, parametric modeling of modular steel frame structures is studied according to their characteristics. An automatic BIM modeling method is developed for modular steel frame structures. Second, the building energy consumption is modeled using various machine learning algorithms based the database that is constructed from the Energyplus software. The proposed XGBoost model provides efficient and accurate predictions for the building energy consumption. Finally, the energy consumption model as well as the cost formula serve as the objective functions in NSGA-Ⅱ algorithm to build the design optimization program for modular steel frame structures. During the optimization, structural bearing capacity must be satisfied. Pareto solution set is then achieved by the developed program and is analyzed. By solving the multi-objective design problem of modular steel frame structures with advanced computing techniques, this study contributes to the intelligent upgrade of the modular steel frame structure industry, and realizes the rapid and efficient design of modular steel frame structures.