diff --git a/vllm/core/scheduler.py b/vllm/core/scheduler.py
index 11d020be0c940..471e540dd4b9b 100644
--- a/vllm/core/scheduler.py
+++ b/vllm/core/scheduler.py
@@ -883,16 +883,21 @@ def _schedule_chunked_prefill(self):
 
         # Update waiting requests.
         self.waiting.extendleft(running_scheduled.preempted)
+
         # Update new running requests.
-        self.running.extend([s.seq_group for s in prefills.seq_groups])
-        self.running.extend(
-            [s.seq_group for s in running_scheduled.decode_seq_groups])
-        self.running.extend(
-            [s.seq_group for s in running_scheduled.prefill_seq_groups])
+        # By default, vLLM scheduler prioritizes prefills.
+        # Once chunked prefill is enabled,
+        # the policy is changed to prioritize decode requests.
         self.running.extend(
             [s.seq_group for s in swapped_in.decode_seq_groups])
         self.running.extend(
             [s.seq_group for s in swapped_in.prefill_seq_groups])
+        self.running.extend(
+            [s.seq_group for s in running_scheduled.decode_seq_groups])
+        self.running.extend(
+            [s.seq_group for s in running_scheduled.prefill_seq_groups])
+        self.running.extend([s.seq_group for s in prefills.seq_groups])
+
         # Update swapped requests.
         self.swapped.extend(running_scheduled.swapped_out)
         return SchedulerOutputs(