Precise modeling of gas injection into a liquid bath is essential for both determining the enormous interaction mechanisms between phases and promoting its better utilization in nuclear industry. For this purpose, an investigation on three CFD methods was performed and compared with the experimental results from air injected vertically into a water pool through a circular nozzle with diameter of 1.58 mm in this work. An inhomogeneous poly-dispersed method based on a bubble-size-dependent lift force model was integrated into the traditional gas-liquid two-fluid model for the simulation of gas-liquid jet flow with air injected vertically in a water bath. Results show that the computer resources consumption can be kept at a low level using the method. Compared with the mono-dispersed and homogeneous poly-dispersed models, the predicted results of the void fraction distribution obtained by using the inhomogeneous poly-dispersed model considering the adjustable coefficients for the break up and coalescence models show good agreements with those of the experiment, with a maximum deviation less than 20%. Hence, the model should be more reliable to predict two-phase flow parameters. Bubble distributions analyses of varying scales were also demonstrated, indicating that the adoption of both inhomogeneous gas flow field and lift force model able to change sign of its coefficient play an important role for predicting the gas injection into liquid bath correctly.