diff --git a/litellm/litellm_core_utils/audio_utils/transcription_streaming.py b/litellm/litellm_core_utils/audio_utils/transcription_streaming.py
index cf2681f5d36..719909d66ea 100644
--- a/litellm/litellm_core_utils/audio_utils/transcription_streaming.py
+++ b/litellm/litellm_core_utils/audio_utils/transcription_streaming.py
@@ -11,8 +11,11 @@ from openai.types.audio import (
     TranscriptionTextDeltaEvent,
     TranscriptionTextDoneEvent,
 )
+from pydantic import TypeAdapter
 
-from litellm.types.utils import TranscriptionResponse
+from litellm.types.utils import TranscriptionDetectedLanguage, TranscriptionResponse
+
+_DETECTED_LANGUAGES: Final = TypeAdapter(list[TranscriptionDetectedLanguage])
 
 
 class TranscriptionStreamLogging(Protocol):
@@ -72,7 +75,11 @@ class _TranscriptionEventCollector:
         response: Final = TranscriptionResponse(
             text=done_event.text if done_event is not None else "".join(self.text_deltas),
             usage=done_event.usage.model_dump() if done_event is not None and done_event.usage is not None else None,
-            languages=([language.model_dump() for language in done_languages] if done_languages is not None else None),
+            languages=(
+                _DETECTED_LANGUAGES.validate_python(done_languages, from_attributes=True)
+                if done_languages is not None
+                else None
+            ),
         )
         if self.duration is not None:
             response.set_audio_transcription_duration(self.duration)
diff --git a/tests/unit/llms/openai/transcriptions/test_gpt_transcribe.py b/tests/unit/llms/openai/transcriptions/test_gpt_transcribe.py
index ccc176e6451..dc8feac29e9 100644
--- a/tests/unit/llms/openai/transcriptions/test_gpt_transcribe.py
+++ b/tests/unit/llms/openai/transcriptions/test_gpt_transcribe.py
@@ -75,7 +75,7 @@ def test_sync_transcription_stream_logs_final_text_and_usage_once() -> None:
             content=(
                 'data: {"type":"transcript.text.delta","delta":"hello "}\n\n'
                 'data: {"type":"transcript.text.done","text":"hello world",'
-                '"usage":{"type":"duration","seconds":2.5}}\n\n'
+                '"usage":{"type":"duration","seconds":2.5},"languages":[{"code":"en"}]}\n\n'
             ),
         )
 
@@ -97,6 +97,7 @@ def test_sync_transcription_stream_logs_final_text_and_usage_once() -> None:
     logged_response: Final = logging_obj.success_handler.call_args.args[0]
     assert logged_response.text == "hello world"
     assert logged_response.usage.model_dump(exclude_none=True) == {"type": "duration", "seconds": 2.5}
+    assert logged_response.model_dump()["languages"] == [{"code": "en"}]
     logging_obj.failure_handler.assert_not_called()