Methane Autothermal Reforming (ATR) and non-catalytic partial oxidation (POX) are two industrial processes used to produce syngas, a mixture of hydrogen and carbon monoxide. In those reactors, methane is burnt with oxygen in fuel-rich conditions. Downstream of the flame, the gaseous combustion products further react with steam and remaining methane in the turbulent "post-flame" region. In order to perform Reynolds Average Navier–Stokes (RANS) simulations of the reactor, accurate modeling strategies are required to compute the average chemical source terms in this post-flame region. In the present study, a DNS numerical experiment has been performed to reproduce the properties of the flow in this part of the reactor. Results are used as a reference to a priori assess the performances of different modeling strategies derived from three turbulent combustion models. The results of this analysis show that, among the three selected models, only the two models based on tabulated chemistry description are able to properly recover the right values of the average chemical source term. The PCM-FPI approach, based on a one-point statistic description using a Beta density probability function, appears as the most accurate approach compared to the two others.