The article examines the word-formation competence of artificial intelligence in translating German feminatives into Ukrainian, with a focus on competing suffix models and compliance with contemporary linguistic norms. The object of the study is feminatives as a component of the Ukrainian word-formation system, while the subject is the translation strategies used by different versions of the ChatGPT language model. The research material includes German official texts containing feminatives denoting professions, positions, and social status. The methodology is based on contrastive and quantitative analysis of German feminatives with the suffix -in/-innen and their Ukrainian equivalents generated by three ChatGPT versions (GPT-4.0, GPT-5.1, and GPT-5.2), correlated with data from the General Ukrainian Language Corpus (HRAK). The results reveal clear differences in word-formation strategies. GPT-5.2 shows the highest sensitivity to modern Ukrainian tendencies, while GPT-5.1 demonstrates a transitional pattern and GPT-4.0 prefers traditional forms. The findings confirm the increasing sensitivity of language models to word-formation variability and corpus-based data.