There are many on this board that know more about this than I do, but the WRF will become operational this fall, and it will replace the current NAM (ETA) model. Given that it's a relatively young model, I'm not sure there's much in the way of verification statistics on how well it handles tropical cyclones. I don't know much beyond that, so hopefully others "in the know" will comment.
Coming from a friend of mine... that worked with an experimental phase of the WRF this past spring - the WRF model performed poorly. Verification numbers were nothing near what they expected and they (SPC) found it to be VERY supercell (convective)-happy. For some reason, it seemed to want to take every single updraft in a conditionally unstable environment and turn it into an HP beast. Newer versions, I'm assuming are performing better and if the operational run hasn't been pushed back to later than this fall, then it must be doing much better than it did during the spring...
Verification/performance archives are also available here...
Granted, this was during the spring... and of course, didn't have any tropical activity to forecast or verify against. So, while most of this is relatively moot, you can get an idea of how it has handled in the past.
Personally, I love the resolution and its capabilities... I've heard it will have the functionality to use satellite and radar data during initialization - which is something very few models have been able to do. So, that alone can/may help accuracy, in at least, the short-term.
Just heard a big talk on NCEP's WRF plans last week at the Great Lakes Op Met Conference in Grand Rapids and it did VERY well with Katrina, also showed some good examples from the Spring convective experiment where the NAM WRF held its own... The PPT should be online soon I would expect.
I like the ETA model as well (though not in tropical situations). The GFS and GFDL are my favorites. For forecasting in general, I find that the No Good Model (NGM) is antique now. Surprised they haven't buried this one yet and replaced it.
I agree to some extent. NGM is excellent at MOS temperature forecasts and also for thunderstorm and severe thunderstorm pops. It is horrible at PTYPE in the northeast from my experience. ETA blows it away on temperature profile in a column of air and on PTYPE.
They're fairly similar, but with slightly different models running... Penn State even has an experimental WRF running locally. The GFDL performs "fairly" well, but is only a tracking model... it doesn't forecast strength very well. The GFDI (the interpolated version of the GFDL) does a much better job at intensity forecasting from what I hear. I found output from it a while back but lost the URL.
The MM5 FSU seems to do fairly well, too. You may want to check that out.