Modern manufacturing value chains require strategic coordination of processes across organizational boundaries to optimize profits while promoting sustainability. However, the adoption of integrated process modeling approaches across value chains faces challenges due to privacy concerns surrounding cross-organizational data sharing. Vertical Federated Learning (VFL) surfaces as a prospective resolution to this predicament, facilitating the joint training of models while preserving the privacy of individual data. Nevertheless, the absence of standardized benchmarks and datasets has so far hindered the progression of research and practical deployment of VFL. In an effort to mitigate this hindrance, we introduce VFLBench, a practical benchmark for VFL focused on smart manufacturing. This benchmark includes a collection of datasets with natural partitions and provides a comprehensive evaluation of state-of-the-art VFL algorithms. Through the establishment of a structured framework for comparative analysis, we aim to stimulate both the research and application of VFL solutions in practical scenarios, particularly within the manufacturing sector.