The abstract definition of temperature is that you have some set of modes (mode = possibly-occupied quantum states with fixed energy) and that there is enough interaction between these modes such that their probability of being occupied has a characteristic dependence on their respective energy. This characteristic dependence is a thermal distribution (which is either Bose-Einstein or Fermi-Dirac statistics).

According to wikipedia (heh), quark-gluon plasma is not really a plasma so much as a liquid. Presumably, that means the modes are still roughly momentum modes, and you can literally think of a hotter QGP as faster quarks and gluons sliding around and slamming into one another. It's a Fermi liquid, though, which means the exclusion principle has significant effect. (I'm a physics grad student, but not a high-energy theorist, so take that for what it's worth.)