Network Digital Twins (NDTs) are emerging as key enablers of intelligent, data-driven automation in next-generation networks. However, existing definitions of NDTs remain generic and overlook telecom-specific requirements such as multi-granular synchronization, modeling fidelity, and integration with Operational and Business Support Systems (OSS/BSS). This paper addresses these gaps by introducing a capability-based taxonomy comprising four classes of NDTs, characterized along key dimensions including purpose, synchronization granularity, modeling detail, and contextual alignment. We further align this taxonomy with functional categories adapted from the TeleManagement (TM) Forum’s Digital Twin for Decision Intelligence (DT4DI) initiative, clarifying how different NDT capabilities support decision intelligence across the 6G lifecycle. Building on this foundation, we present a standards-compliant architectural framework grounded in the TM Forum Open Digital Architecture (ODA) to enable deployment of diverse NDTs through OSS/BSS platforms. The paper’s contributions are primarily conceptual, encompassing the proposed taxonomy, architectural framework, and representative 6G use cases, and are complemented by a survey that synthesizes recent research, industry developments, and quantitative economic assessments of NDT adoption, all aimed at informing academic and industry stakeholders on the state of the art and the key challenges associated with scalable and sustainable NDT deployment.