Eight atmospheric reanalyses were compared against observed vertical profiles of temperature, specific humidity and wind speed collected by two research aircraft in February–March 2010 in the Antarctic Peninsula region. These data offered a rare possibility to validate reanalyses against independent in situ data which have not been assimilated into the reanalyses. The reanalyses had generally too moist profiles with too low wind speeds, but otherwise the errors in the reanalyses had large spatial differences. On the eastern side of the peninsula, the near-surface temperatures were largely overestimated. None of the reanalyses outperformed the others in all variables, at all altitudes and on both sides of the peninsula. Generally, NCEP-CFSR and MERRA had the smallest errors in temperature profiles, JRA-55 had marginally the most accurate specific humidity profiles andNCEP-CFSR had the best wind profiles. The reanalyses were coherent, although biased, on the western side of the Antarctic Peninsula, but on the eastern side the spread was large. All the reanalyses underestimated the variability between the individual profiles of temperature and wind speed. The modern reanalyses with a sufficient spatial resolution and an adequate data assimilation method outperformed the others, especially on the eastern side.