Text this: Benchmarking Large Language Models for News Summarization