diff --git a/litellm/responses/litellm_completion_transformation/transformation.py b/litellm/responses/litellm_completion_transformation/transformation.py
index df298f7c448..c3add946014 100644
--- a/litellm/responses/litellm_completion_transformation/transformation.py
+++ b/litellm/responses/litellm_completion_transformation/transformation.py
@@ -1794,6 +1794,12 @@ class LiteLLMCompletionResponsesConfig:
             ):
                 output_details_dict["text_tokens"] = completion_details.text_tokens
 
+            if (
+                hasattr(completion_details, "image_tokens")
+                and completion_details.image_tokens is not None
+            ):
+                output_details_dict["image_tokens"] = completion_details.image_tokens
+
             if output_details_dict:
                 response_usage.output_tokens_details = OutputTokensDetails(
                     **output_details_dict
diff --git a/tests/test_litellm/responses/litellm_completion_transformation/test_litellm_completion_responses.py b/tests/test_litellm/responses/litellm_completion_transformation/test_litellm_completion_responses.py
index 5074bbf4397..b183875d050 100644
--- a/tests/test_litellm/responses/litellm_completion_transformation/test_litellm_completion_responses.py
+++ b/tests/test_litellm/responses/litellm_completion_transformation/test_litellm_completion_responses.py
@@ -1424,6 +1424,47 @@ class TestUsageTransformation:
         assert response_usage.input_tokens_details is None
         assert response_usage.output_tokens_details is None
 
+    def test_transform_usage_with_image_tokens(self):
+        """Test that image_tokens from Vertex AI/Gemini are properly transformed to output_tokens_details"""
+        # Setup: Simulate Vertex AI/Gemini usage with image_tokens in completion_tokens_details
+        usage = Usage(
+            prompt_tokens=10,
+            completion_tokens=150,
+            total_tokens=160,
+            completion_tokens_details=CompletionTokensDetailsWrapper(
+                reasoning_tokens=0,
+                text_tokens=50,
+                image_tokens=100,  # From Vertex AI candidatesTokensDetails with modality="IMAGE"
+            ),
+        )
+
+        chat_completion_response = ModelResponse(
+            id="test-response-id",
+            created=1234567890,
+            model="gemini-2.0-flash",
+            object="chat.completion",
+            usage=usage,
+            choices=[
+                Choices(
+                    finish_reason="stop",
+                    index=0,
+                    message=Message(content="Here is the generated image.", role="assistant"),
+                )
+            ],
+        )
+
+        # Execute
+        response_usage = LiteLLMCompletionResponsesConfig._transform_chat_completion_usage_to_responses_usage(
+            chat_completion_response=chat_completion_response
+        )
+
+        # Assert
+        assert response_usage.output_tokens == 150
+        assert response_usage.output_tokens_details is not None
+        assert response_usage.output_tokens_details.reasoning_tokens == 0
+        assert response_usage.output_tokens_details.text_tokens == 50
+        assert response_usage.output_tokens_details.image_tokens == 100
+
 
 class TestStreamingIDConsistency:
     """Test cases for consistent IDs across streaming events (issue #14962)"""