Catalytic performance of ceria-supported metals can be dramatically improved when the support is in the form of nanoparticles rather than bulky samples. To rationalize this finding, ceria nanoparticle models are required that are tractable at a high computational level albeit adequately represent surface properties of larger moieties dealt with experimentally. This work addresses the issue of smallest ceria nanoparticles, for which adsorption parameters become sufficiently independent of the particle size. For that, adsorption of a CO probe molecule on a particular surface site of a series of model nanoparticles containing similar to 50-230 atoms has been calculated using the DFT + U approach. (c) 2008 Elsevier B. V. All rights reserved.