You can get to it at http://moe.met.fsu.edu/mm5/ as it runs; the runs later get input into http://moe.met.fsu.edu/tcgengifs/ with some additional/different products there. The model does reasonably well with the overall pattern & its features, but there hasn't been a real system out there since we changed the convective scheme in the model about 10 days to two weeks ago (Irene occurred during the middle of that). Thus, we don't know entirely how well it is going to perform from here on out. My guess is that intensity will be underdone, but it's a necessary trade-off to keep terrain features from playing a huge role.

Generally, it is run every 6hr when there is a tropical storm in the basin and every 12hr when there is not one. Technical glitches sometimes interrupt the runs.