We present an extremely efficient and rather general model in which whole polymer chains are represented as soft particles. The particles are characterized by their overall sizes and shapes, as given by the conformations of the underlying chains. The probability of occurrence of a particle with a given size determines its internal free energy. The density of monomers within each particle is calculated from all conformations that have the same size. The interaction between two particles is taken to be proportional to the spatial overlap of their monomer density distributions. When a large number of such particles are brought into contact, as is the case for a polymer melt, the interactions between the particles force them to shrink and modify the equilibrium size distribution. We show by simulations that this model leads to a Gaussian statistics of the chains in melt. Since the internal degrees of freedom of a chain are integrated out, a large number (of order 10(4)) of long (e.g., N=100) chains can be simulated within seasonable computer time on a single work-station processor. A straightforward extension of this model is used to study symmetric polymer blends.