Rayleigh-Benard (RB) convection is a challenging topic for both academic research and for several applications in the field of atmospheric physics, industrial applications, and building insulation. The actual applications involve high Rayleigh or Grashoff numbers but the Direct Numerical Simulations (DNS) remain limited to relatively small values of these numbers. Approximate models are required for the prediction of flow fields and heat transfer at higher numbers. The aim of this study is to assess the ability of a Large Eddy Simulation (LES) model to predict accurate mean values and second order moments of temperature and velocity fields. The reference fields are provided by DNS simulations under the Boussinesq approximation in a RB cell filled with air at a Rayleigh number (Ra) up to 10⁹. The employed numerical DNS method is based on a Tchebychev pseudo-spectral method and the most appropriate LES model in this case appears to be the Spectral Vanishing Viscosity (SVV) model where high order modes are exponentially damped to stabilize the numerical scheme. The predicted mean velocity and temperature fields from this LES model, as well as their second order moments, are compared to DNS values. A sensitivity study to the SVV parameters, i.e. the Tchebychev mode cutoff and the damping amplitude, is carried out. The model appears to be a promising approach for accurate predictions at much higher Rayleigh numbers.