The article is devoted to the problems of neural-machine translation, discusses the degree of its quality and, especially, the adequacy of special vocabulary translation (in the field of oil and gas industry). The article starts with a short description of the history of machine translation and its types. Special attention is paid to neural-machine translation, its basic characteristics and criteria of translation quality evaluation. Also the authors mention the types of mistakes that arise in the process of translation. The next part of the article contains examples of comparison of quality of translation of articles devoted to oil and gas industry. The translations were produced by Yandex and Google translators. The authors come to the conclusion that both systems generally provide an adequate quality of translation, and existing mistakes are mostly connected with translation of oil and field terminology, homonymous to terminology of other field of science, when the machine translator prefers a more frequently used term. Summing up, the authors state the importance of increasing the quality of teaching the machine translator, which needs a growing number of published glossaries of terms in various fields of science, experts' assistance and an increased role of human translator's involvement in the field of science, where he/she works.