This paper reports on a project to compare the representation of the semiannual oscillation (SAO) in the equatorial stratosphere and lower mesosphere among six major global atmospheric reanalysis datasets and with recent satellite SABER and MLS observations. All reanalyses have a good representation of the quasi-biennial oscillation (QBO) in the equatorial lower and middle stratosphere and each displays a clear SAO centered near the stratopause. However, the differences among reanalyses are much more substantial in the SAO region than in the QBO dominated region. The degree of disagreement among the reanalyses is characterized by the standard deviation (SD) of the monthly-mean zonal wind and temperature; this depends on latitude, longitude, height, and time. The zonal wind SD displays a prominent equatorial maximum that increases with height, while the temperature SD is minimum near the equator and largest in the polar regions. Along the equator the zonal wind SD is smallest around the longitude of Singapore where consistently high-quality near-equatorial radiosonde observations are available. Interestingly the near-Singapore minimum in SD is evident to at least ~3 hPa, i.e. considerably higher than the usual ~10 hPa ceiling for in situ radiosonde observations. Our measurement of the agreement among the reanalyses shows systematic improvement over the period considered (1980–2016), up to near the stratopause. Characteristics of the SAO at 1 hPa, such as its detailed time variation and the displacement off the equator of the zonal wind SAO amplitude maximum, differ significantly among the reanalyses. Disagreement among the reanalyses becomes still greater above 1 hPa. One of the reanalyses in our study also has a version produced without assimilating satellite observations and a comparison of the SAO in these two versions demonstrates the very great importance of satellite derived temperatures in the realistic analysis of the tropical upper stratospheric circulation.