Summarization of textual content is an essential task in natural language processing (NLP), enabling efficient information retrieval and comprehension. Traditional summarization models, while effective, often produce outputs that can miss critical information or introduce redundancy. This paper proposes a novel approach to enhance the quality of summaries by combining outputs from multiple summarization models and refining them using semantic similarity measures. Specifically, we utilize the finetuned BART, T5, and PEGASUS models to generate initial summaries. These outputs are then combined and redundant sentences are removed through semantic similarity analysis using the Sentence-BERT embeddings. Our experimental results demonstrate that this multi-model approach produces more comprehensive and coherent summaries, capturing a broader range of information with reduced redundancy. This method offers significant improvements in the quality of text summarization, making it a valuable tool for applications requiring precise and concise summaries.