Prompted by the Yohkoh observations of solar flares, which have established the essential role of magnetic reconnection in the release of energy, we have studied the evolution of eruptive flares in some detail based on the reconnection model by means of the two-dimensional magnetohydrodynamic (MHD) simulations. We are interested in what factor affects the time evolution of solar flares and how the related phenomena, particularly observed loop-top source and plasmoid eruption, can be explained by this model. We have studied the dependence of the structure and evolution of the system on plasma beta (ratio of gas pressure to magnetic pressure), adiabatic index, gamma, and rho(c) (initial density in the current sheet). If the timescale and velocity are normalized by Alfven time and Alfven speed, respectively, we find that the main results (e.g., reconnection rate, plasmoid velocity, etc.) are rather insensitive to the plasma beta. The rho(c) value, on the other hand, crucially affects the motion of a plasmoid: the ejection velocity of plasmoid grows in proportion to rho(c)(-1/2) in the early phase, which suggests that the observed plasmoid velocity can be reproduced when we assign rho(c) similar or equal to 40 rho(0) (initial density outside the current sheet). When adiabatic index gamma is small, corresponding to the case of efficient thermal conduction, plasma heating will be generally suppressed, but the compression effect can be rather enhanced, which plays an important role in forming the high-density loop-top source. We discuss loop-top sources, plasmoid eruption, and the rise motion of a loop in comparison with the observations. Our simulations can well account for the existence of the loop-top, hard X-ray sources discovered in the impulsive flares. We concluded that both the impulsive flares and the LDE (long duration event) flares can be generally understood by the reconnection model for the cusp-type flares.