Language translation makes it possible for people to communicate, share information, and build strong relations. Because it converts text from one language to another, Neural Machine Translation contributes to the development of the performances. For the English–Assamese language pair, this paper analyses four Neural Machine Translation models based on distinct Techniques. This paper presents four models of Sequence Architecture, which include: (1) Bidirectional Gated recurrent units (GRUs) (2) Bidirectional Long Short-Term Memory (BI-LSTM) (3) Bidirectional Gated recurrent units (GRUs) +BPE (4) Bidirectional Long Short-Term Memory (BI-LSTM) +BPE, with attention mechanism. The results of the comparison revealed that GRU with BPE outperforms Bi-LSTM in terms of performance.