diff --git a/src/agents/voice/models/openai_stt.py b/src/agents/voice/models/openai_stt.py
index b1f1b6da7..733406f04 100644
--- a/src/agents/voice/models/openai_stt.py
+++ b/src/agents/voice/models/openai_stt.py
@@ -226,7 +226,10 @@ async def _handle_events(self) -> None:
                     break
 
                 event_type = event.get("type", "unknown")
-                if event_type == "input_audio_transcription_completed":
+                if event_type in [
+                    "input_audio_transcription_completed",  # legacy
+                    "conversation.item.input_audio_transcription.completed",
+                ]:
                     transcript = cast(str, event.get("transcript", ""))
                     if len(transcript) > 0:
                         self._end_turn(transcript)
diff --git a/tests/voice/test_openai_stt.py b/tests/voice/test_openai_stt.py
index ecc41f2e2..f1ec04fdc 100644
--- a/tests/voice/test_openai_stt.py
+++ b/tests/voice/test_openai_stt.py
@@ -184,22 +184,35 @@ async def test_stream_audio_sends_correct_json():
 
 
 @pytest.mark.asyncio
-async def test_transcription_event_puts_output_in_queue():
+@pytest.mark.parametrize(
+    "created,updated,completed",
+    [
+        (
+            {"type": "transcription_session.created"},
+            {"type": "transcription_session.updated"},
+            {"type": "input_audio_transcription_completed", "transcript": "Hello world!"},
+        ),
+        (
+            {"type": "session.created"},
+            {"type": "session.updated"},
+            {
+                "type": "conversation.item.input_audio_transcription.completed",
+                "transcript": "Hello world!",
+            },
+        ),
+    ],
+)
+async def test_transcription_event_puts_output_in_queue(created, updated, completed):
     """
-    Test that a 'input_audio_transcription_completed' event
+    Test that a 'input_audio_transcription_completed' event and
+    'conversation.item.input_audio_transcription.completed'
     yields a transcript from transcribe_turns().
     """
     mock_ws = create_mock_websocket(
         [
-            json.dumps({"type": "transcription_session.created"}),
-            json.dumps({"type": "transcription_session.updated"}),
-            # Once configured, we mock a completed transcription event:
-            json.dumps(
-                {
-                    "type": "input_audio_transcription_completed",
-                    "transcript": "Hello world!",
-                }
-            ),
+            json.dumps(created),
+            json.dumps(updated),
+            json.dumps(completed),
         ]
     )