The present study investigates the performance of dynamic mixed model (DMM; Zang et ai.) in a channel with strong wall injection through a Large eddy simulation (LES) technique. The DMM results are compared with those of DNS and the results obtained with popular dynamic Smagorinsky model (DSM). Better agreement is achieved when using the DMM with box filter than DSM and coarse DNS in predicting the first and second order statistics as well as large-scale structures of velocity and temperature fields. Such favorable features of DMM are attributed to the fact that it explicitly calculates the modified Leonard stress term and only models the remaining cross and the SGS Reynolds stress terms and, thus, it reduces the excessive burden put on the model coefficient of DSM. Also it is demonstrated that the DMM can be successfully extended to the prediction of temperature (passive scalar) field where strong streamwise inhomogeneity exists.