To address the limitations of traditional models in cross-laboratory validation, this study proposes a chloride diffusion coefficient model for concrete that incorporates a cement type factor, accounting for the influence of cement type and strength grade. The model is developed through regression analysis of multi-source large-sample Rapid Chloride Migration (RCM) test data. Firstly, a comprehensive database of 179 RCM test datasets from 70 laboratories was established to analyze the effects of water-binder ratio, cement type, and strength grade on the chloride diffusion coefficient via regression. Furthermore, the cement type factor was introduced into the computational model using a two-phase regression method, and its value was determined based on the multi-source large-sample data. Finally, comparative analyses with traditional models and validation using independent test data were conducted. The results show that the proposed multi-source large-sample model improves the fitting accuracy to experimental data by 19.6%compared to conventional mono-source small-sample models. The cement type factor effectively captures the combined influence of cement type and strength, reducing the weighted average error and coefficient of variation by 32.0% and 25.0%, respectively, thereby significantly enhancing the model's predictive precision and adaptability.