This study was aimed at discovering the levels of English-Arabic code-mixing in Islamic boarding school students’ daily conversation. This study departs from code-mixing theory in sociolinguistics. Qualitative approach was utilized as the research design. The data were in the form of word, phrase, and clause taken from utterances obtained by recording the conversations of students. From the data analysis taken from 20 situations, there were 98 utterances of code mixing, which were used by students in their conversation at Islamic boarding school covering intra- and extra-sentential code – mixing. The dominant occurrence is in word level totalling to 65 occurrences (66.4%), followed by clause level 13 occurrences (13.2%), phrase level 10 occurrences (10.2%), and sentence level 10 occurrences (10.2%). Therefore, the code – mixing did exist in students’ conversation. Even though the appearance of the code – mixing in students’ conversation was not really high but this needs to be re-examined the correct use of the code – mixing in orther to avoid the misunderstanding and miscommunication among the students and with teachers.