Abstract: To meet the increasing demand of numerical reactor analysis for efficient, reliable, and physically consistent two-phase thermal-hydraulic models, a reactor channel-scale two-phase flow solver, YoGuang-two fluids (YG-TF), was developed in this work on the basis of the finite-volume computational platform OpenFOAM. Conventional subchannel codes generally relied on a 1.5-dimensional modeling strategy, in which the axial flow was predominantly resolved, whereas transverse transport and local three-dimensional effects were treated through simplified empirical or semi-empirical closures. Although this approach has been widely used in engineering analysis, its capability to describe complex flow redistribution, local phase interaction, and multidimensional two-phase behavior remains limited. In contrast, YG-TF adopted a non-equilibrium two-fluid formulation, in which the conservation equations of the liquid and vapor phases were solved separately within an interpenetrating-continuum framework. This modeling strategy enabled the solver to resolve realistic three-dimensional geometries and dominant flow features at the reactor channel scale, while maintaining a computational cost comparable to that of conventional subchannel methods. The single-phase and two-phase thermal-hydraulic performance of YG-TF was systematically assessed using representative validation problems, including a flow split test case, the Bartolomei subcooled boiling experiments, and the BFBT international benchmark rod-bundle experiments. The flow split case was employed to examine the capability of the solver to capture lateral mass transfer induced by transverse pressure differences, which was a key phenomenon in rod-bundle and channel-scale reactor thermal-hydraulic analysis. The Bartolomei experiments were used to evaluate the prediction of subcooled boiling and void fraction development under heated-channel conditions. The BFBT benchmark further provides a rod-bundle-scale assessment of outlet void fraction and pressure-drop characteristics under more realistic reactor-relevant geometrical and operating conditions. The validation results show that YG-TF can effectively predict flow redistribution driven by transverse pressure gradients. For the Bartolomei subcooled boiling experiments, the mean absolute deviation of the calculated void fraction is less than 0.035, indicating satisfactory agreement with the experimental measurements. For the BFBT rod-bundle experiments, the maximum absolute deviation of the outlet void fraction is 0.048, and the mean relative deviation of the pressure drop does not exceed 11%. These results demonstrate that YG-TF provides reliable predictive capability for both single-phase redistribution and two-phase boiling flow characteristics at the reactor channel scale. The present work therefore provides an efficient numerical tool for channel-scale two-phase thermal-hydraulic analysis of nuclear reactors and establishes a technical foundation for the development of independently developed and controllable numerical reactor analysis capabilities.