diff --git a/litellm/responses/utils.py b/litellm/responses/utils.py
index ee0f1c6338f..a35aaac6b3c 100644
--- a/litellm/responses/utils.py
+++ b/litellm/responses/utils.py
@@ -958,13 +958,23 @@ class ResponseAPILoggingUtils:
             response_api_usage, "output_tokens_details", None
         )
         if output_tokens_details:
-            completion_tokens_details = CompletionTokensDetailsWrapper(
-                reasoning_tokens=getattr(
-                    output_tokens_details, "reasoning_tokens", None
-                ),
-                image_tokens=getattr(output_tokens_details, "image_tokens", None),
-                text_tokens=getattr(output_tokens_details, "text_tokens", None),
-            )
+            if isinstance(output_tokens_details, dict):
+                completion_tokens_details = CompletionTokensDetailsWrapper(
+                    **output_tokens_details
+                )
+            else:
+                completion_tokens_details = CompletionTokensDetailsWrapper(
+                    reasoning_tokens=getattr(
+                        output_tokens_details, "reasoning_tokens", None
+                    ),
+                    image_tokens=getattr(
+                        output_tokens_details, "image_tokens", None
+                    ),
+                    text_tokens=getattr(output_tokens_details, "text_tokens", None),
+                    audio_tokens=getattr(
+                        output_tokens_details, "audio_tokens", None
+                    ),
+                )
 
         chat_usage = Usage(
             prompt_tokens=prompt_tokens,
diff --git a/tests/test_litellm/responses/test_responses_utils.py b/tests/test_litellm/responses/test_responses_utils.py
index 2b3e601b134..bd441321507 100644
--- a/tests/test_litellm/responses/test_responses_utils.py
+++ b/tests/test_litellm/responses/test_responses_utils.py
@@ -327,7 +327,8 @@ class TestResponseAPILoggingUtils:
             "output_tokens_details": {
                 "reasoning_tokens": 30,
                 "image_tokens": 100,
-                "text_tokens": 70,
+                "text_tokens": 50,
+                "audio_tokens": 20,
             },
         }
 
@@ -346,7 +347,8 @@ class TestResponseAPILoggingUtils:
         assert result.completion_tokens_details is not None
         assert result.completion_tokens_details.reasoning_tokens == 30
         assert result.completion_tokens_details.image_tokens == 100
-        assert result.completion_tokens_details.text_tokens == 70
+        assert result.completion_tokens_details.text_tokens == 50
+        assert result.completion_tokens_details.audio_tokens == 20
 
     def test_transform_response_api_usage_with_realtime_keys(self):
         """Realtime input_token_details / output_token_details normalize for Usage."""
@@ -375,6 +377,7 @@ class TestResponseAPILoggingUtils:
 
         assert result.completion_tokens_details is not None
         assert result.completion_tokens_details.text_tokens == 12
+        assert result.completion_tokens_details.audio_tokens == 8
 
     def test_transform_response_api_usage_tokens_details_keep_values(self):
         """Keeps input_tokens_details / output_tokens_details when singular keys are also present."""