Machine translation (MT) has revolutionized communication, breaking down language barriers. Traditional approaches like statistical machine translation (SMT) have achieved significant success, but contemporary upgrade in large language models (LLMs) extends a favorable alternative. An equivalent study of modern LLMs for MT tasks. We evaluate and compare their performance on various language pairs using established metrics to assess translation quality, fluency, domain-specificity, and potential biases. By analyzing these strengths and weaknesses, we aim to deliver discernment into the effectiveness of different LLM models for MT. This research will guide researchers and practitioners in nominating the most satisfactory LLM for specific translation needs, furthering the development of LLM-based MT and its impact on the field.