This paper presents an innovative approach that combines MarkBERT-BiLSTM with contextual CRF modeling and is designed to address key challenges in Chinese named entity recognition (NER). These challenges include the difficulty in recognizing subtle features, the complexity involved in feature extraction, ambiguous lexical boundaries and the intricate nature of contextual semantics. The method commences with the utilization of the MarkBERT model for preprocessing, which generates vectors that convey information regarding lexical boundaries. These pretrained vectors are subsequently fed into a BiLSTM framework, wherein the global context mechanism facilitates the integration of both forward and backward sentence information into the BiLSTM structure in an effective manner. Ultimately, a CRF layer is employed. At the entity labeling and classification stage, the generation of the optimal predicted sequence is ensured. The experimental results indicate that the model performs exceptionally well on the MSRA dataset, achieving a high F1 score of 95.75%. This outcome highlights the significant advantages of the proposed model in addressing the complexities of Chinese NER.