Large eddy simulations of flame extinction with N2 as extinguish agent are performed focusing on combustion and radiation modelling with infinitely fast chemistry. The use of a EDC combustion model with dynamically determined coefficients, an enthalpy-based flame extinction model based on a locally variable critical flame temperature and the use of the WSGGM for radiation are employed in order to predict flame extinction in a turbulent CH4 line burner. The numerical predictions of mean temperatures, combustion efficiencies and radiative fractions with different grid sizes are compared to the experiments by White et al. (2015).Overall, the results from the numerical simulations agree well qualitatively and, to some extent, quantitatively with the experimental data when small grid sizes are employed. More specifically, the maximum values and the profile widths of the mean temperatures at two axial locations examined are reasonably well predicted. The decrease in the combustion efficiencies as the extinction limit is approached is reproduced by the numerical simulations. The decreasing trend in the radiative fractions as the oxidizer stream is diluted with N2 is also captured by the simulations using a WSGGM model for radiation with a dynamically determined beam length which is calculated based on the local heat release rate. Nevertheless, the resulting radiative fractions are still over-predicted as the extinction limit is approached. Limitations of some of the typically used approaches regarding radiation modelling in flame extinction scenarios are outlined. Additionally, possible deficiencies on the use of a fixed critical flame temperature and/or re-ignition temperature in the flame extinction/re-ignition modelling are discussed and possible extensions of the modelling approaches are presented.