Higher order modulation means with each use of the channel, you transmit more bits. For example, in $$8-$$ary QAM, 3 bits are transmitted in each channel use. In other words, each 3 bits are mapped to one of the 8 available waveforms to be transmitted over the channel. If you use $$16-$$ary QAM, then 4 bits are transmitted with each channel use. In general, for $$M-$$ary QAM, $$\log_2(M)$$ bits are transmitted with channel use. The larger $$M$$, the more bits are transmitted per channel use. However, as $$M$$ increases, the distance between the constellation points decreases, and this means that the probability of error increases. So, there should be a trade-off between the constellation size, and the probability of error.