This paper presents our approach to the PolEval 2024 Task 1 on Polish language reading comprehension, utilizing state-of-the-art Large Language Models (LLMs). We developed a system that effectively handles both answer generation and answerability classification by leveraging decoder-only models. Our solution addresses key challenges including processing long contexts exceeding typical model limitations and identifying questions that cannot be answered from the given text. The system achieves strong performance on both the Levenshtein score for answer quality and the F1 score for answerability classification.