The article considers the use of recurrent neural networks for handwriting recognition in Kazakh and Russian languages based on Cyrillic graphics. Handwritten text recognition is a complex structured process consisting of scanning paper with text, dividing the received information into texts and images, using methods of intelligent recognition of manuscripts and processing the results. In this scientific work, the problem of recognition of Kazakh and Russian handwritten text was solved using a new deep neural network model based on a fully closed CNN proposed by Abdallah, and the results were analyzed. The paper describes a model based on the Gated-CNN-BGRU architecture (CCN, convolutional neural networks, Bidirectional gated recurrent unit), and calculates the error rate of characters, the error rate of words and the error rate of sentences for recognizing handwritten texts. Russian and Kazakh handwritten data sets HKR (Handwritten Kazakh & Russian) and KOHTD (Kazakh Offline Handwritten Text Dataset) in Kazakh and Russian were used for training and testing of manuscript recognition systems. The proposed model was implemented using the TensorFlow library for Python.