Intrusion detection is an important research topic in cyber-security. In intrusion detection problem we need to distinguish between malicious and benign network traffic. In recent years, multiple approaches have been designed to tackle the problem. Usually the features are hand picked or trained using an encoder model. With the recent development of large language models (LLMs), in this paper we evaluate the the embedding technique of LLMs for intrusion detection problem. The experimental results showed that the performance can reached to more than 99% of accuracy in CIC-IDS-2017 dataset.