This paper is based on the process of building a language model using the Word2Vec tool on articles from the Croatian Wikipedia. In the introductory section, various types of language models and their role in natural language processing are explored. Focusing on the Word2Vec tool, its mechanism of distributed word representation based on semantic similarities among words in context is analyzed. The practical part of the work describes the steps involved in constructing a language model using the Word2Vec tool. The process included text cleaning and preparation, tokenization, and context creation for each word. Through iterative updates of vector representations, the model acquired semantic features of words, which were confirmed through the analysis of distance and word-analogy metrics. The results obtained by applying Word2Vec to articles from the Croatian Wikipedia demonstrated that the model successfully captures semantic relationships between words, contributing to the understanding of linguistic structures. This work provides a foundation for further research in the field of language models and emphasizes the importance of the Word2Vec tool in the process of building and developing language resources. Through this paper, it is clear that the application of Word2Vec enables progress in understanding linguistic patterns, thereby facilitating better utilization of language in various applications. Such research opens doors to further innovations in the field of natural language processing, encouraging development in how we communicate and use language in the modern digital environment.