An approach for modeling differential diffusion effects in turbulent nonreacting and reacting jets with stochastic mixing models is proposed. This approach has been applied to the modified Curl's mixing model and the Interaction-by-Exchange-with-the-Mean (IEM) mixing model used in the joint scalar Probability Density Function (PDF) approach. For comparison, a more elaborate mixing model based on the Linear-Eddy Model (LEM) has been incorporated into the PDF approach. Numerical results are assessed by comparisons with experimental data by Smith et al. (1995a, b) with a fuel mixture of 36%H-2 and 64%CO2 and the data by Meier et al. (1996b) with a fuel mixture of 50%H-2 and 50%N-2. The predicted and measured differential diffusion effects are examined extensively with reference to computed laminar opposed flow flames. For nonreacting jets, the numerical results and data agree reasonably well. However, less agreement is found in the numerical results and the jet flame data obtained by Smith et al. (1995b). For the 50%N-2+50%H-2 turbulent jet flames, the numerical results agree with the data better at locations far downstream than in the near field. The observed discrepancies may be due to the re-laminarization of the how in the near field. Moreover, results obtained with the modified Curl's and the IEM model with differential diffusion effects are in good agreement with those obtained with the Linear-Eddy Model which is computationally expensive for high Reynolds number flows.