What is the difference between agent mode and chat mode aside from having image recognition capabilities?

As in chat mode, we can also use the functions capability and hence “agentic”, what is the benefit of switching to agent mode if we don’t need image recognition?