The 8-class QoS model is the appropriate DiffServ design when an enterprise is adding voice and video to an existing data network and needs priority treatment for voice. Cisco enterprise QoS designs commonly evolve from a simple data-only model to a multi-class model that separates routing/control, voice bearer, video, signaling, transactional data, bulk data, scavenger traffic, and default traffic. This separation is important because voice requires strict low latency, low jitter, and minimal loss, while video usually requires guaranteed bandwidth but must not starve voice or network control traffic. A 4-class model can be too coarse when both VoIP and IP video are introduced because it often collapses distinct traffic types into broad queues. A 12-class model may be appropriate for very mature or complex environments, but it is more operationally demanding.
The 6-class model can work in some cases, but the common Cisco campus and WAN baseline for mixed voice, video, and data is the 8-class model. Reference topics: DiffServ QoS, enterprise QoS class models, voice priority queue, video classification, DSCP design.