Due to the inherent challenges in dataset construction, no comprehensive Chinese Chemical Named Entity Recognition (CNER) dataset has been publicly released to date. As a result, there is a scarcity of domain-specific Chinese datasets in the field of chemistry, impeding progress in key intelligent chemistry tasks such as reaction mechanism analysis, automatic literature classification, and cross-lingual terminology alignment. This work addresses this gap by introducing the first large-scale Chinese CNER dataset—cnChemNER—which leverages granted Chinese patents as a data source to develop a rule-based, patent-oriented largely automated extraction framework. The dataset comprises 76,245 annotated chemical entities spanning four primary categories, which are further divided into 73 fine-grained subcategories. The experimental results validate the effectiveness of using patents as a data source, with language models trained on cnChemNER demonstrating superior performance across precision, recall, and F1-score metrics. cnChemNER bridges a critical gap in Chinese chemical corpora and represents a significant advancement in terms of annotation granularity, semantic structure, and domain-specific adaptability. We anticipate that cnChemNER will serve as a valuable resource for enhancing model performance in cross-lingual chemical text processing. Additionally, it can support downstream tasks in related vertical domains such as biomedicine and materials science.