Smurfjet,
before you start pulling out "raw" data (its actually very refined... you don't want to see the really raw data
) and using it operationally, I would recommend you take some time to research the various numerical models you are using. Find out how they work, what data they are using, how often runs are done, resolution and many other characteristics of numerical models. You need to know what you are looking at (just like you need to learn how airborne wx radar works before using operationally for storm avaidance).
Otherwise you might be very unpleasantly surprised with you get.
To answer your question as to where to look, you first need to tell us at in what region of the world you are looking at, how far into the future you need to be able to get the data, what altitudes and with what kind of spatial and temporal resolution?
In some cases having only pictorial representaions with wind barbs (and their related uncertainty in determining exact wind direction) can be more representative of the actual uncertainties in forecasted winds aloft. Giving you "exact" numerical answers might mislead you into believing that there are less uncertainties then there really are.
palgia
http://www.nco.ncep.noaa.gov/pmb/products/