Abstract: Viscosity is a physical property of fluids and shows resistance to flow. In metallurgical slag, it directly affects various parameters of a metallurgical process such as reaction rate, separation effect, etc. The estimation of viscosity by models during a production process is considered to be much more effective owing to the production fluctuations and complexities of the slag composition. Many viscosity models have been developed in the past, such as the structural model with a wide range of adaptability and complex calculation process and the empirical and semiempirical models with simple structure and a narrow range of adaptability. The present paper proposed a new method to calculate the structural parameter (NBO/T) ratio (the amount of nonbridging oxygen per tetrahedral-coordinated atom), based on which the relationship between the viscosity of molten slag and (NBO/T) ratio was investigated. First, the viscosity model was applied to SiO₂–ΣM_xO slags, with the model parameters obtained by fitting the viscosity data of pure oxide and SiO₂–M_xO binary slag, and the average deviations were in the range of 9%–18.5%. Then, the model was extended to calculate the viscosity of SiO₂–Al₂O₃–ΣM_xO, a multicomponent complex aluminosilicate system, and Al₂O₃ was split into acid and basic oxides. Then the oxides were used for calculating the (NBO/T) ratio and viscosity activation energy based on the amphoteric behavior of Al₂O₃ in SiO₂–Al₂O₃–M_xO ternary slag system. Using the parameters of a SiO₂–M_xO binary system, the model parameters of the Al₂O₃-containing slag system were obtained by fitting the viscosity data of the SiO₂–Al₂O₃–M_xO ternary slag system with average deviations between 10% and 25%. In addition, the viscosity of a multi-complex system (SiO₂–Al₂O₃–CaO–MgO–FeO–Na₂O–K₂O–Li₂O–BaO–SrO–MnO) and its subsystems were also calculated using the model proposed in this paper, and the average deviations is less than 25%, which shows relatively accurate prediction results. The present model is based on the analysis of a slag structure and processing of data by an empirical method. This method has a better prediction effect and wider application range compared with the traditional empirical model, and it uses a simpler calculation process compared with the structure model.