diff --git a/python/pyspark/sql/connect/dataframe.py b/python/pyspark/sql/connect/dataframe.py
index 4552fbe18f40..6600cd0d739f 100644
--- a/python/pyspark/sql/connect/dataframe.py
+++ b/python/pyspark/sql/connect/dataframe.py
@@ -2075,6 +2075,7 @@ def is_cached(self) -> bool:
 
     def toLocalIterator(self, prefetchPartitions: bool = False) -> Iterator[Row]:
         query = self._plan.to_proto(self._session.client)
+        binary_as_bytes = self._get_binary_as_bytes()
 
         schema: Optional[StructType] = None
         for schema_or_table in self._session.client.to_table_as_iterator(
@@ -2089,7 +2090,7 @@ def toLocalIterator(self, prefetchPartitions: bool = False) -> Iterator[Row]:
                 if schema is None:
                     schema = from_arrow_schema(table.schema, prefer_timestamp_ntz=True)
                 yield from ArrowTableToRowsConversion.convert(
-                    table, schema, binary_as_bytes=self._get_binary_as_bytes()
+                    table, schema, binary_as_bytes=binary_as_bytes
                 )
 
     def pandas_api(