Gibbs ensemble Monte Carlo simulations and cloud point measurements were performed to understand the molecular weight dependence of χ and the effect of dispersity on the phase behavior of polymer mixtures. Oligomeric blends consisting of poly(ethylene-alt-propylene) (PEP) and poly(ethylene oxide) dimethyl ether (PEO) were used as the model systems. First, the molecular weight dependence of χ for PEP/PEO mixtures was studied using simulations and experiments for PEP/PEO mixtures with various molecular weights. An empirical model with a single adjustable parameter kij is used to quantify this molecular weight dependence, and it allows for the accurate prediction of χ of PEP/PEO mixtures with arbitrary molecular weights. Second, the effects of molecular weight distribution (MWD) and dispersity (Đ) of PEO on the PEP/PEO phase diagram were investigated via both simulations and experiments. When PEO is relatively monodisperse (Đ < 1.2), the phase diagram is found to be insensitive to either MWD or Đ, despite differentiation in molecular partitioning observed from simulations. However, the coexistence curve for mixtures containing PEO with a bimodal distribution and a large dispersity (Đ = 1.76) differs dramatically from that for mixtures containing low-dispersity PEO, which suggests that the former mixture can no longer be treated as a binary system. Furthermore, structural analysis was performed from simulation trajectories to probe microscopic heterogeneity and aggregation behavior in the liquid phases. The results in this work permit the accurate prediction of χ and the phase diagram of disperse binary polymeric mixtures.