The statistical properties of the flow structure in the Yucatan Channel's main section based on 2 years of continuous observations (August 1999 to June 2001) are compared with those simulated by the Oc¿an Parall¿lis¿ (OPA) primitive equation model in two different configurations: the ATL6 configuration from the CLIPPER project (the whole Atlantic domain with a 1/6¿ resolution), and the Prototype Atlantique M¿diterran¿e (PAM) configuration from the MERCATOR project (the North Atlantic domain with 1/12¿ resolution). While the observed 2-year mean transport into the Gulf of Mexico is 23.06 Sv, ATL6 has a 5-year mean of 27.49 Sv and PAM has a 3-year mean of 29.06 Sv. Apart from this discrepancy in the mean transport both simulations are able to reproduce well, in structure and magnitude, details of the mean current and its variability. Features like the Yucatan Current core and the deep countercurrents in the Yucatan and Cuban sides are both well reproduced, as is the presence of a minimum surface current variability at the center of the channel. The two principal empirical modes of current variability, which are related to the passage of eddies through the channel, are also reproduced by the simulations. The observed and simulated transports above and below the 6¿C isotherm show no coherency at any frequency except for a significant coherent peak at a period around 20 days, which we interpret as a first baroclinic Kelvin mode between the two layers. As in the observations, the shedding of eddies by the Loop Current in the simulations is preceded by periods of diminished negative horizontal vorticity flux into the Gulf of Mexico.