Here's the analysis by Jeff Masters.

All of the major models agree that Ike will hit eastern Cuba on Sunday night. After this point, the models continue to diverge. A southern camp of models, the ECMWF and UKMET, take Ike across eastern Cuba and into the western Caribbean, then across the western tip of Cuba or Yucatan Channel between Cuba and Mexico's Yucatan Peninsula. These models predict an eventual landfall near the Mexico/Texas border a week from now. This track would bring tropical storm conditions to the Cancun/Cozumel area beginning Tuesday afternoon or evening, with possible hurricane conditions by Wednesday morning.

The northern camp of models, including the GFS, NOGAPS, GFDL, and HWRF, turn Ike west-northwest over Cuba, forecasting that Ike will pop off the coast of Cuba near the Florida Keys on Tuesday, then swing north-northwest. The northward turn is delayed in the current runs, putting Ike into the central Gulf of Mexico, or several hundred miles offshore the western Florida coast. The trough of pressure that pulls Ike to the north is expected to be weak, leaving Ike in a region of weak steering currents. A similar situation occurred in 1985, when Category 3 Hurricane Elena got stranded in the Gulf and wander offshore of the Florida Panhandle for several days. So far, the GFDL has done a good job with Ike, so I will continue to lean towards that track. The GFDL tracks Ike over Cuba until the storm pops off the coast south of the Keys, and intensifies it from a borderline Category 1 or 2 hurricane to a Category 3 hurricane as it passes 50 miles southwest of Key West. The GFDL brings Category 2 winds to Key West. The model then takes Ike north-northwestward into the Gulf of Mexico to an uncertain future.