Background: This study examines the evolution of English speaking proficiency test methods, which have traditionally relied on face-to-face interactions to assess communicative language competence. Recently, computer-based language tests have also been used on a larger scale, albeit with concerns about their impact on measurement. Objectives: This paper investigates the potential of the metaverse in the context of Oral Proficiency Interviews (OPI) for evaluating the speaking proficiency of English learners. Specifically, the research centres on the utilisation of one of the recently emerging metaverse platforms (Zepeto) to conduct OPI tests based on the comparison of the results obtained from the traditional face-to-face tests and an online video conferencing platform (Zoom). Methods: A total of 60 EFL young pupils participated in three different speaking tasks via the three modes given in a counter-balanced order. The test scores of the three modes were rated and analysed to examine the degree to which different test modes had an effect on them. A questionnaire with 10 items (plus one sub-question) was used to investigate participants' experiences and perceptions of three speaking assessment modes. Results and Conclusions: The results show that test-takers performed significantly better in the metaverse test mode than in the other two modes, while no significant difference was found between the face-to-face and online video conferencing modes. The sub-scores of speaking test criteria such as task completion, pronunciation, language use and fluency were also analysed for comparison, and similar results were found except for the sub-scores of the language use criteria. Responses to the questionnaire also showed that the metaverse platform was the most helpful and preferred mode. Implications for Practice: Based on the findings, this study suggests that the new testing modes incorporated with virtual environments have great potential to be alternatives to traditional testing modes.