Cybersecurity knowledge graph is an important means to conduct deep correlation analysis on unstructured Cyber Threat Intelligence （CTI） data. The Use of Large Language Model （LLM） can overcome the limitations of traditional cybersecurity named entity recognition methods， such as lack of understanding of text context， difficulty in capturing long-distance dependencies， and lack of robustness when facing noise and redundancy. However， the extraction of cybersecurity named entities by a single LLM will lead to poor adaptability of the model capability outside the domain boundary， insufficient generalization ability on Out-Of-Distribution （OOD） data， and cognitive limitations of a single model. To solve these problems， an LLM collaborative linking method that integrates fine-tuning was proposed. In the method， Low-Rank Adaptation （LoRA） fine-tuning of LLMs was performed on a base LLM using a CTI corpus， and the model's understanding of the cybersecurity field was enhanced through parameter optimization. At the same time， static voting and dynamic meta-learner linking were adopted by LLM collaboration. In the static mechanism， prediction results of multiple models were integrated through voting； and in the dynamic mechanism， the feature space adaptation function was designed to quantify the model capability boundary. Simultaneously， dynamic mapping from input data features to model performance was achieved through meta-learners such as random forests. Experimental results show that the F1-score of the proposed method reaches 91.1%， which is 6.8% higher than that of the semantic enhancement network method based on multiple features， providing an efficient and scalable solution for named entity recognition in cybersecurity.