The increasing complexity of cyberthreats has made effective network intrusion detection systems (NIDS) crucial. Traditional NIDS, which rely on predefined signatures or normal network behavior, often struggle with high false-positive rates and emerging threats. This study explores integrating Large Language Models (LLMs) into NIDS to improve detection accuracy and adaptability. Fine-tuned on a comprehensive NetFlow dataset, the LLMs were evaluated using accuracy, precision, recall, and F1 score. The results demonstrate LLM-based NIDS’ potential in reducing false positives and improving novel attack detection, marking a promising direction for cybersecurity.