(AB) Significantly higher efficiency, greater power output (typically 3X or more) and an implied requirement for local or global feedback to linearize the switching transition in the output devices.
@lynn_olson I think most people haven't thought this bit through. If there is a switching transition in the output devices (producing crossover distortion) then the amp is actually biased class C.
Class B is defined as exactly 1/2 of the audio waveform. If its more than that the output device is biased AB, if less than that its class C. Class B has really only been theoretical on this account until perhaps recently when microcontrollers could be used to actually really get the output device(s) to do exactly 50% of the waveform.
EV used to sell amps back in the 1950s they claimed were class B, such as the A20 or A30. Indeed they made double the power you would expect out of a pair of the same power tubes in class A and the tubes ran cool. They had no zero crossing artifact at any power level you could measure (I have an A20 so I've put it through its paces)...
This all says they were really very lightly biased AB amplifiers. AB amps won't have a zero crossing artifact unless there is a design flaw (usually a problem in the driver circuit). Feedback won't fix a zero crossing artifact since there is no loop gain at the zero crossing if the outputs are in cutoff. So if the amp is AB there's no worries since there's loop gain.
Put simply, most of the time when people talk about a zero crossing artifact such as 'switching transition' they are talking about an amp that is biased class C. Douglas Self wrote about an amp he designed called the 'blameless amplifier', which is an example of a class C push-pull amp, although he claimed it was class B. But if you really think about it, you know the class B in the real world can't exist without computer control.