Comparing LLMs and NMTs Performances in Translating English Indonesian Texts

As global communication increases, the demand for effective translation tools grows, particularly in countries with low English proficiency such as Indonesia. This study compares Neural Machine Translation (NMT) systems, Google Translate and DeepL, with Large Language Models (LLMs), ChatGPT and Gemini, in translating English–Indonesian and Indonesian–English texts. Using theFLORES-200 and BPPT-PANL corpora, we evaluated outputs with BLEU, chrF, and COMET metrics. While LLMs demonstrate contextual fluency and achieve a leading score in one of eight evaluation scenarios, NMT systems consistently outperform across the remaining cases in both corpora. These results affirm the current superiority of NMT systems in delivering reliable and structurally accurate translations, while LLMs show emerging potential in specific, context-heavy tasks.