Introduction (relevance). The development of a language model for pharmaceuticals using artificial intelligence is one of the most significant challenges in modern pharmaceutical information. Currently, the pharmaceutical industry faces several significant challenges, including the difficulty of selecting the optimal medication, the need to quickly process large amounts of information, and the increasing demands for patient care. Modern AI systems have the potential to significantly improve the efficiency of medication selection, provide accurate recommendations, and minimize the risk of errors. However, the successful implementation of such information technologies is impossible without a clear and structured approach. This is where the need for the development of specialized programs arises, taking into account the specific features of the pharmaceutical market, the range of medicines, and the needs of professionals and patients. The development of such a model becomes an important step in the implementation of innovative technologies in pharmacies, with the aim of improving the quality of pharmaceutical services, reducing the risks of side effects, and preventing the misuse of medications. Purpose: To review and analyze the process of preparing initial data and training a prototype of a neural network-based software solution that aims to provide requested pharmaceutical information about drugs. Materials and methods. The study used content, marketing, and structural analysis methods, as well as questionnaires, comparison, generalization, systematization, and modeling. The study used the following information sources: the MedInfo medical reference and information system, the Vidal drug reference, and the State Register of Medicines of the Russian Federation (GRЛС) (both editions dated 2025). Results. The conducted comparative analysis showed the positive aspects of using software for conducting pharmaceutical business and the demand, first of all, for mobile, easily adaptable complexes using AI. Taking into account all the advantages and disadvantages of the existing software complexes, a model of information support was proposed, which correctly determined the presence and absence of information on specific drugs, adequately compared data on indications, contraindications, and side effects, and demonstrated the ability to make elementary generalizations within the given database. In addition, in the absence of information about the drug, the model also correctly identified the limited data and did not generate false information, which confirms the correctness of the chosen principle of system instructions and the limitation of logical assumptions. Discussions. The conducted research has confirmed the possibility of using local language models in pharmaceutical practice as a tool for intellectual support of specialists, which will optimize the consulting process, increase the accuracy of information support, and automate a number of routine operations in pharmacy activities. Conclusions (conclusion). The developed prototype of local language models demonstrated high efficiency in use in pharmaceutical practice, confirming the prospects of introducing intelligent tools, thanks to which, the quality of informing patients, the accuracy of recommendations and interpretations of requests from pharmacy visitors, is minimized, the risk of erroneous conclusions is reduced. Thus, the foundation is laid for the development of innovative solutions based on AI, which contributes to improving the level of pharmaceutical assistance.