When should I use pyspark saveAsTable?

In pyspark I can save using saveAsTable():

  df = spark.read.parquet('input') \
    .write \
    .format('parquet') \
    .saveAsTable('output')

or without it:

  df = spark.read.parquet('input') \
    .write \
    .parquet('output')

What is the difference between them?