Abstract: Critical heat flux (CHF) is a crucial safety limit for PWR core design. Accurately predicting CHF in the core rod bundle channel can enhance the economy and safety of the reactor. Currently, the prediction of rod bundle CHF in engineering is mainly based on specific fuel assembly experiments and subchannel code modeling to obtain local parameters that can be used to develop the corresponding CHF correlation. However, the traditional correlation of CHF of the rod bundle cannot be extrapolated effectively due to the rough division of subchannels and the inaccuracy of local parameter calculation of the rod surface. The CHF correlation developed for specific fuel assemblies can only be applied to specific rod bundle geometries. The rod bundle structure of the guide tube requires additional punishment, which does not conform to the local phenomenon hypothesis and cannot achieve real prediction. To predict CHF at different rod bundle geometries, such as new fuel assemblies and new reactors, a mechanism model of CHF of the rod bundle was established based on the high-precision subchannel program developed by Xi’an Jiaotong University. The highprecision subchannel code is further subdivided based on the conventional subchannel division and uses the griddistributed resistance model. Two types of gaps formed by new subchannels are treated separately using the turbulent mixing model. The high-precision subchannel code makes the flow field calculation more accurate. Based on Weisman’s bubble crowding model, a bundle CHF mechanism model was proposed. The turbulent intensity distribution near the bundle wall was obtained using the concept of equivalent tube. The roughness of the bubble layer thickness was considered, and the bubble diameter at the CHF position was solved iteratively. The developed bundle CHF mechanism model was embedded into the high-precision subchannel code and predicted using the bundle CHF experimental database of Xi’an Jiaotong University. The operation condition range of verified data covers the typical operating condition range of PWR, including two rod bundle forms of typical lattice and guide tube lattice, with a total of 601 CHF data points. The assessment results demonstrate that the average ratio of the predicted value to the experimental value is 0.99, and the mean square error is 4.69%, and the deviation has no obvious parameter bias with pressure, mass flux, and critical quality. Furthermore, the model can be well applied to the prediction of CHF of typical lattice and guide tube lattice. The guide tube lattice CHF can be predicted successfully without any modification of the CHF model. This model reveals the essence of the cold wall effect at the mechanism level.