Canada’s new short-term greenhouse gas (GHG) emissions target, resulting from the 2015 Paris Agreement, has led the energy production industry to adopt more carbon-neutral fuels, including biomass. Hence, the development of biomass combustion technology has recently gained more attention owing to its capability to burn a wide range of biomass fuels. Computer simulation of biomass furnaces is a very important step towards improving the combustion performance and emissions of these power generation systems. Combustion models play a key role in the reliability of the numerical simulation of the gas-phase combustion of biomass combustion systems. The aim of the present numerical study is to evaluate the prediction capability of flamelet-based partially premixed combustion models in simulating the gas-phase combustion process of a grate-firing biomass furnace. Additionally, the effects of the adopted premixed models (i.e., extended coherent flame model and C-equation based model) and non-premixed models (i.e., steady diffusion flamelet (SFM) and unsteady diffusion flamelet (UFM)) on the overall prediction of the partially premixed model are assessed. The predicted temperature field and species concentrations are compared with published experimental measurements and also with published numerical simulations which use other combustion models (i.e., EDC/Flamelet hybrid model, SFM and UFM). The results of this study reveal that except for the slow-forming and chemically dominated species, partially premixed combustion models (both extended coherent flame model/SFM and C-equation/SFM-based partially premixed model) are capable of reproducing the experimental temperature and major species with reasonable accuracy and low computational expense. C-equation/UFM-based partially premixed model is found to be the most optimum combination amongst all examined partially premixed models for overcoming the deficiency faced while predicting the slow-forming species.