Abstract: The development, verification, and validation of reactor core physics design software packages are critical in pressurized water reactor (PWR) nuclear power plant design and represent a key research area in reactor physics. TORCH is a core physics calculation software package developed by Nuclear Power Institute of China (NPIC). In the assembly calculation, based on the multi-group library, the subgroup resonance method was employed to handle the resonance self-shielding effect, the method of characteristics (MOC) transport method was used to obtain the multi-group assembly flux spectrum, and the predictor-corrector depletion method was applied to generate the homogenized multi-group cross-sections at each burnup point. Combining base mode correction calculations, branch calculations, and parametric fitting, the homogenized few-group parameter library for the assembly was produced. In the core calculation, the few-group parameters were obtained through interpolation based on burnup, control rod position, and boron concentration. The traditional Green’s function diffusion method was used to solve the neutron diffusion equation, and a single-channel model was employed to account for thermal-hydraulic feedback, thereby obtaining key core physics parameters such as critical boron concentration and power distribution at each burnup point. In this study, the two-step algorithm and key theoretical models were introduced, including the 45-group multigroup library based on ENDF/B-Ⅵ, subgroup resonance method, GPU-accelerated MOC transport method, predictor-corrector microscopic depletion method, equivalent one-dimensional reflector modeling method, Green’s function expansion nodal method, and detector reaction rate calculation method. Validation was conducted using M310 and HPR1000 PWR designs, covering startup physics tests and power operation data from Fangjiashan, Fuqing, Zhangzhou, and Karachi nuclear power plants. The key parameters for evaluation included critical boron concentration, isothermal temperature coefficient, control rod worth, and detector reaction rate. For startup tests, critical boron concentration analysis based on 61 data points shows a root mean square deviation of 17 ppm and a maximum deviation of 50 ppm. The isothermal temperature coefficient evaluation of 40 data points yields a root mean square deviation of 0.73% and a maximum relative deviation of 1.53%. Control rod worth assessment based on 231 data points shows a root mean square deviation of 4.12%, with 7 instances exceeding the 10% engineering criterion, all occurring at positions with low rod worth. In power operation analysis, examination of 312 burnup points for critical boron concentration reveals a root mean square deviation of 21 ppm and a maximum deviation of 62 ppm, where 7 cases surpass the 50 ppm engineering criterion at either the beginning or end of the fuel cycle. Detector reaction rate measurements display a maximum relative deviation of 2.10%. These results confirm TORCH’s high computational accuracy against measured data, supporting its engineering application in PWRs.