The performance of solution processed polymer:fullerene thin film photovoltaic cells is largely determined by the nanoscopic and mesoscopic morphology of these blends that is formed during the drying of the layer. Although blend morphologies have been studied in detail using a variety of microscopic, spectroscopic, and scattering techniques and a large degree of control has been obtained, the current understanding of the processes involved is limited. Hence, predicting the optimized processing conditions and the corresponding device performance remains a challenge. We present an experimental and modeling study on blends of a small band gap diketopyrrolopyrrole-quinquethiophene alternating copolymer (PDPP5T) and [6,6]-phenyl-C71-butyric acid methyl ester ([70]PCBM) cast from chloroform solution. The model uses the homogeneous Flory-Huggins free energy of the multicomponent blend and accounts for interfacial interactions between (locally) separated phases, based on physical properties of the polymer, fullerene, and solvent. We show that the spinodal liquid-liquid demixing that occurs during drying is responsible for the observed morphologies. The model predicts an increasing feature size and decreasing fullerene concentration in the polymer matrix with increasing drying time in accordance with experimental observations and device performance. The results represent a first step toward a predictive model for morphology formation.