I guess we could probably search the Katrina threads to find out for sure. I thought it was the GFDL. I recall it pointing to NO while the rest still wanted her to hit Florida, but you might be right.
I tend to give the GFDL more weight than the other models because it was designed for tropical cyclones. But it hasn't been consistent yet with Rita.
Another run of the models later that afternoon shows LBAR predicting a second lanfall at Venice, LA and then again on the MS/AL border. Other models also called the MS/AL border for second landfall.
Later that same day LBAR shifts left again with second landfall at Terrebonne Parish, then through NOLA, across the lake and into MS.
An early morning run on the 26th has LBAR as a far western outlier with second landfall at St. Mary Parish and through Lafayette.
Five hours later on the 26th LBAR is nearly spot on with the center of the storm passing to the east of NOLA. All other models were calling Pensacola and Panama City, FL.
Late on the 26th everybody shifts west with GDFL being right on the money for the future strike.
By mid-day on the 28th pretty much every model was converging on the NOLA hit. UKMET and NHC still wanted Moss Point/Pascagoula, MS. UKMET held out till early morning on the 29th to converge.