diff --git a/litellm/llms/custom_httpx/llm_http_handler.py b/litellm/llms/custom_httpx/llm_http_handler.py
index 4c9abaad908..928c76c8b42 100644
--- a/litellm/llms/custom_httpx/llm_http_handler.py
+++ b/litellm/llms/custom_httpx/llm_http_handler.py
@@ -2598,10 +2598,12 @@ class BaseLLMHTTPHandler:
                 provider_config=responses_api_provider_config,
             )
 
-        return responses_api_provider_config.transform_get_response_api_response(
+        result = responses_api_provider_config.transform_get_response_api_response(
             raw_response=response,
             logging_obj=logging_obj,
         )
+        result._hidden_params["response_cost"] = 0.0
+        return result
 
     async def async_get_responses(
         self,
@@ -2673,10 +2675,12 @@ class BaseLLMHTTPHandler:
                 provider_config=responses_api_provider_config,
             )
 
-        return responses_api_provider_config.transform_get_response_api_response(
+        result = responses_api_provider_config.transform_get_response_api_response(
             raw_response=response,
             logging_obj=logging_obj,
         )
+        result._hidden_params["response_cost"] = 0.0
+        return result
 
     #####################################################################
     ################ LIST RESPONSES INPUT ITEMS HANDLER ###########################
@@ -4495,9 +4499,9 @@ class BaseLLMHTTPHandler:
                         # Second: Execute agentic loop
                         # Add custom_llm_provider to kwargs so the agentic loop can reconstruct the full model name
                         kwargs_with_provider = kwargs.copy() if kwargs else {}
-                        kwargs_with_provider[
-                            "custom_llm_provider"
-                        ] = custom_llm_provider
+                        kwargs_with_provider["custom_llm_provider"] = (
+                            custom_llm_provider
+                        )
                         agentic_response = await callback.async_run_agentic_loop(
                             tools=tool_calls,
                             model=model,
@@ -4613,9 +4617,9 @@ class BaseLLMHTTPHandler:
                         # Second: Execute agentic loop
                         # Add custom_llm_provider to kwargs so the agentic loop can reconstruct the full model name
                         kwargs_with_provider = kwargs.copy() if kwargs else {}
-                        kwargs_with_provider[
-                            "custom_llm_provider"
-                        ] = custom_llm_provider
+                        kwargs_with_provider["custom_llm_provider"] = (
+                            custom_llm_provider
+                        )
                         agentic_response = (
                             await callback.async_run_chat_completion_agentic_loop(
                                 tools=tool_calls,
@@ -5099,7 +5103,10 @@ class BaseLLMHTTPHandler:
         _is_async: bool = False,
         fake_stream: bool = False,
         litellm_metadata: Optional[Dict[str, Any]] = None,
-    ) -> Union[ImageResponse, Coroutine[Any, Any, ImageResponse],]:
+    ) -> Union[
+        ImageResponse,
+        Coroutine[Any, Any, ImageResponse],
+    ]:
         """
 
         Handles image edit requests.
@@ -5311,7 +5318,10 @@ class BaseLLMHTTPHandler:
         fake_stream: bool = False,
         litellm_metadata: Optional[Dict[str, Any]] = None,
         api_key: Optional[str] = None,
-    ) -> Union[ImageResponse, Coroutine[Any, Any, ImageResponse],]:
+    ) -> Union[
+        ImageResponse,
+        Coroutine[Any, Any, ImageResponse],
+    ]:
         """
         Handles image generation requests.
         When _is_async=True, returns a coroutine instead of making the call directly.
@@ -5551,7 +5561,10 @@ class BaseLLMHTTPHandler:
         fake_stream: bool = False,
         litellm_metadata: Optional[Dict[str, Any]] = None,
         api_key: Optional[str] = None,
-    ) -> Union[VideoObject, Coroutine[Any, Any, VideoObject],]:
+    ) -> Union[
+        VideoObject,
+        Coroutine[Any, Any, VideoObject],
+    ]:
         """
         Handles video generation requests.
         When _is_async=True, returns a coroutine instead of making the call directly.
diff --git a/tests/test_litellm/llms/openai/responses/test_openai_responses_transformation.py b/tests/test_litellm/llms/openai/responses/test_openai_responses_transformation.py
index 802868aa7a9..c9abd2e00f8 100644
--- a/tests/test_litellm/llms/openai/responses/test_openai_responses_transformation.py
+++ b/tests/test_litellm/llms/openai/responses/test_openai_responses_transformation.py
@@ -1282,3 +1282,82 @@ class TestPhaseParameter:
         assert validated[0]["phase"] == "commentary"
         assert validated[1]["phase"] == "final_answer"
         assert "phase" not in validated[2]
+
+
+class TestGetResponsesZeroCost:
+    """Test that GET /v1/responses/{response_id} does not incur duplicate costs (bug #25015)."""
+
+    def _make_fake_response(self):
+        """Create a fake httpx response with usage data (as OpenAI would return)."""
+        fake_response_json = {
+            "id": "resp_abc123",
+            "object": "response",
+            "created_at": 1700000000,
+            "status": "completed",
+            "model": "o3-deep-research",
+            "output": [],
+            "usage": {
+                "input_tokens": 100,
+                "output_tokens": 50,
+                "total_tokens": 150,
+            },
+        }
+        return httpx.Response(
+            status_code=200,
+            json=fake_response_json,
+            request=httpx.Request("GET", "https://api.openai.com/v1/responses/resp_abc123"),
+        )
+
+    def test_get_responses_sets_zero_cost(self):
+        """
+        Verify that get_responses (sync) sets response_cost=0.0 in _hidden_params,
+        so the cost pipeline does not re-bill for a retrieval request.
+        """
+        from litellm.llms.custom_httpx.http_handler import HTTPHandler
+        from litellm.llms.custom_httpx.llm_http_handler import BaseLLMHTTPHandler
+
+        handler = BaseLLMHTTPHandler()
+        config = OpenAIResponsesAPIConfig()
+        logging_obj = MagicMock()
+        litellm_params = GenericLiteLLMParams(api_base="https://api.openai.com/v1/responses")
+
+        mock_client = MagicMock(spec=HTTPHandler)
+        mock_client.get.return_value = self._make_fake_response()
+
+        result = handler.get_responses(
+            response_id="resp_abc123",
+            responses_api_provider_config=config,
+            litellm_params=litellm_params,
+            logging_obj=logging_obj,
+            client=mock_client,
+        )
+
+        assert result._hidden_params["response_cost"] == 0.0
+
+    @pytest.mark.asyncio
+    async def test_async_get_responses_sets_zero_cost(self):
+        """
+        Verify that async_get_responses sets response_cost=0.0 in _hidden_params,
+        so the cost pipeline does not re-bill for a retrieval request.
+        """
+        from litellm.llms.custom_httpx.http_handler import AsyncHTTPHandler
+        from litellm.llms.custom_httpx.llm_http_handler import BaseLLMHTTPHandler
+
+        handler = BaseLLMHTTPHandler()
+        config = OpenAIResponsesAPIConfig()
+        logging_obj = MagicMock()
+        litellm_params = GenericLiteLLMParams(api_base="https://api.openai.com/v1/responses")
+
+        mock_client = AsyncMock(spec=AsyncHTTPHandler)
+        mock_client.get.return_value = self._make_fake_response()
+
+        result = await handler.async_get_responses(
+            response_id="resp_abc123",
+            responses_api_provider_config=config,
+            litellm_params=litellm_params,
+            logging_obj=logging_obj,
+            client=mock_client,
+            custom_llm_provider="openai",
+        )
+
+        assert result._hidden_params["response_cost"] == 0.0