The use of nonaqueous solvents for conversion of biomass molecules into platform chemicals and fuels is hindered by a lack of solubility data and models. Accordingly, the solubility of the model carbohydrate, cellobiose, was measured at 20 degrees C in 18 organic liquids and water. Solubility ranged from <0.001 g L-1 in hydrocarbon solvents to >100 g L-1 in hydrogen bond acceptor solvents. Three models were analyzed to evaluate accuracy and to understand factors that affect cellobiose solubility: Hansen's solubility theory (HST), linear free energy relationship (LFER), and several versions of the UNIQUAC functional-group activity coefficients (UNIFAC) model. Cellobiose Hansen solubility parameters (HSPs) were estimated using experimental solubility measurements, but the model was unable to correctly distinguish several of the good solvents from the poor ones. A regressed LFER containing contributions from solvent excess molar refraction, polarizability/dipolarity, acidity, basicity, and molar volume was able to match measured solubilities to within +/- 2 log units (AAD of 66% and ARD of 27%). The LFER model indicated that good solvents were highly polarizable and had low molar volume, consistent with the good cellobiose solubility observed for water. A modified version of UNIFAC, including an additional association term (A-UNIFAC), predicted the solubility of cellobiose and several other carbohydrates in water and alcohols to within +/- 0.65 log units (AAD of 25% and ARD of 6.7%), indicating that the A-UNIFAC framework can be used in a predictive manner. This work provides guidance for future studies on the use of nonaqueous solvents for biomass conversion.