In this study, we compare modeled (MIT-CAM3) and observed (CARIBIC) sub-micrometer nucleation (N4-12, 4 <= d <= 12 nm) and Aitken mode (N-12, d > 12 nm) particle number concentrations in the upper troposphere and lowermost stratosphere (UT/LMS). Modeled and observed global median N4-12 and N-12 agree fairly well (within a factor of two) indicating that the relatively simplified binary H2SO4-H2O nucleation parameterization applied in the model produces reasonable results in the UT/LMS. However, a comparison of the spatiotemporal distribution of sub-micrometer particles displays a number of discrepancies between MIT-CAM3 and CARIBIC data: N4-12 is underestimated by the model in the tropics and overestimated in the extra-topics. N-12 is in general overestimated by the model, in particular in the tropics and during summer months. The modeled seasonal variability of N4-12 is in poor agreement with CARIBIC data whereas it agrees rather well for N-12. Modeled particle frequency distributions are in general narrower than the observed ones. The model biases indicate an insufficient diffusive mixing in MIT-CAM3 and a too large vertical transport of carbonaceous aerosols. The overestimated transport is most likely caused by the constant supersaturation threshold applied in the model for the activation of particles into cloud droplets. The annually constant SO2 emissions in the model may also partly explain the poor representation of the N4-12 seasonal cycle. Comparing the MIT-CAM3 with CARIBIC data, it is also clear that care has to be taken regarding the representativeness of the measurement data and the time frequency of the model output.