This study focuses on the model reduction of a two-phase loop thermosyphon. The aim is to propose a nonlinear reduced order model able to mimic the thermo-hydraulic behavior of the loop in order to use it for real-time state feedback control, in future applications. First, the one-dimensional two-phase flow model describing the liquid-gas mixture in both mechanical and thermal equilibrium is recalled. The numerical resolution of this detailed model is carried out using a finite volume approach and a Harten-Lax-van Leer Contact Riemann solver. Then, from this detailed model, a new structure of reduced model is determined via the Galerkin projection method. These reduced models, built by the Modal Identification Method, show a very good agreement between the outputs of the detailed model and those computed by the reduced model, during the identification stage. Two test cases, corresponding to different thermal loads at the evaporator, show that the overall levels of density, velocity, mass flow rate, pressure, temperature and internal energy in the loop are satisfactorily reproduced by the reduced model with a global relative error less than 5%. The interest of using such a model lies in the significant gain in CPU time. (C) 2018 Elsevier Ltd. All rights reserved.