A Versatility Analysis: Investigating Large Language Models' Performance Beyond Conventional Benchmarks