In a class B amplifier, both halves of the push-pull output stage are cut off at 0 input. As the input increases, the conducting half starts to conduct in a highly non-linear manner as the input increases. This is "crossover distortion". As the input increases, the conducting half starts to behave in a more linear fashion with respect to the input. To get around the crossover distortion problem, both halves of the output stage are biased so that they are conducting with zero input. This is "Class AB" operation. The higher the bias current, the lower the distortion. The price you pay for the lower distortion is higher power dissipation and lower efficiency. If you increase the bias current to the point where both output stages are conducting a current that is equivalent to the current that would flow at 1/2 output, this is "Class A" operation. In class A operation, the dissipation in the output stage is highest, and is independent of the output.
"Class AB" is so named because it is a compromise between Class A (low distortion, high dissipation) and "Class B" (high distortion, low dissipation)
One of the dangers of class AB or class A operation is the possibility of thermal runaway. This is a condition where, as the output transistors or FETs heat up, the bias current tends to increase, causing a furhter increase in temperature.......Some kind of negative feed back mechanism, such as emitter/source resistors is usually used to mitigate the thermal runaway tendency.
By varying the bias current in a class AB circuit, you can change the compromise between distortion and dissipation.