Abstract: The 3D temperature fluctuation phenomenon caused by the mixing of the coaxial-jet hot and cold fluids was simulated by Fluent software. Several special turbulence models were applied to prediction of this phenomenon, i.e. large eddy simulation model (LES), Reynolds stress model (RSM) and standard k-ε model. By the comparison of the computed data and experimental ones, it is shown that LES is capable of predicting the mixing process. LES model best predicts the time-averaged temperature in the radius, height and azimuth directions. Reynolds averaged Navier-Stokes method (RANS) predicts the extended mixing of the hot and cold fluids. It is also shown that the transient temperature fluctuations are accurately predicted by LES model, while those not by RANS.