To improve the computational speed of complex numerical models and achieve parallel coupled computations across multiple programs, this study develops a distributed coupled computational method based on the Visual Interactive Transient Analysis Code for nuclear Reactor System (VITARS) developed by Xi'an Jiao tong University. This approach allows for external coupling by enabling data exchange between processes via interfaces, transferring parameters like pressure, liquid-phase specific enthalpy, vapor-phase specific enthalpy, void fraction, non-condensable gas fraction, and flow velocity. The computational progress across processes is programmatically synchronized to maintain consistency at interaction moments. Using this method, the platform supports both multi-process parallel computation within a single program and parallel computation across multiple programs. Furthermore, three different distributed splitting methods are proposed based on varying boundary conditions. A series of component-level mechanism tests and analyses of the computation results from splitting a complete two-loop thermal–hydraulic model are conducted to explore the effects of applying these methods on the platform. The results indicate that this method achieves distributed coupled computations between processes with high accuracy and significantly enhances computational efficiency.