GPT-4o Understands Text, But Does It See Clearly? A Benchmarking Study of MFMs on Vision Tasks