From 8a051733cbd46133343a5bf40869c0044fcf4420 Mon Sep 17 00:00:00 2001
From: Naineel Soyantar <naineel.soyantar@zuru.com>
Date: Mon, 7 Sep 2026 10:04:37 +0530
Subject: [PATCH] fix(openrouter): track Responses API stream cost

---
 .../openrouter/responses/transformation.py    | 31 ++++++++++++++++++-
 ...est_openrouter_responses_transformation.py | 30 ++++++++++++++++++
 2 files changed, 60 insertions(+), 1 deletion(-)

diff --git a/litellm/llms/openrouter/responses/transformation.py b/litellm/llms/openrouter/responses/transformation.py
index 30303ac0e4c..5c3b0a2ab3e 100644
--- a/litellm/llms/openrouter/responses/transformation.py
+++ b/litellm/llms/openrouter/responses/transformation.py
@@ -15,7 +15,12 @@ import httpx
 import litellm
 from litellm.llms.openai.responses.transformation import OpenAIResponsesAPIConfig
 from litellm.secret_managers.main import get_secret_str
-from litellm.types.llms.openai import ResponseInputParam, ResponsesAPIResponse
+from litellm.types.llms.openai import (
+    ResponseCompletedEvent,
+    ResponseInputParam,
+    ResponsesAPIResponse,
+    ResponsesAPIStreamingResponse,
+)
 from litellm.types.router import GenericLiteLLMParams
 from litellm.types.utils import LlmProviders
 
@@ -143,3 +148,27 @@ class OpenRouterResponsesAPIConfig(OpenAIResponsesAPIConfig):
             pass
 
         return response
+
+    def transform_streaming_response(
+        self,
+        model: str,
+        parsed_chunk: dict,
+        logging_obj: Any,
+    ) -> ResponsesAPIStreamingResponse:
+        response_event: Final = super().transform_streaming_response(
+            model=model,
+            parsed_chunk=parsed_chunk,
+            logging_obj=logging_obj,
+        )
+        if not isinstance(response_event, ResponseCompletedEvent):
+            return response_event
+
+        usage: Final = response_event.response.usage
+        if usage is None or usage.cost is None:
+            return response_event
+
+        response_event.response._hidden_params["additional_headers"] = {
+            **response_event.response._hidden_params.get("additional_headers", {}),
+            "llm_provider-x-litellm-response-cost": float(usage.cost),
+        }
+        return response_event
diff --git a/tests/test_litellm/llms/openrouter/responses/test_openrouter_responses_transformation.py b/tests/test_litellm/llms/openrouter/responses/test_openrouter_responses_transformation.py
index cb577686a09..43424e9e5f8 100644
--- a/tests/test_litellm/llms/openrouter/responses/test_openrouter_responses_transformation.py
+++ b/tests/test_litellm/llms/openrouter/responses/test_openrouter_responses_transformation.py
@@ -186,6 +186,36 @@ class TestOpenRouterResponsesAPICostTracking:
             "additional_headers", {}
         )
 
+    def test_transform_streaming_response_extracts_completed_response_cost(self):
+        config = OpenRouterResponsesAPIConfig()
+
+        result = config.transform_streaming_response(
+            model="openai/gpt-5-mini",
+            parsed_chunk={
+                "type": "response.completed",
+                "response": {
+                    "id": "resp_abc123",
+                    "object": "response",
+                    "created_at": 1700000000,
+                    "status": "completed",
+                    "model": "openai/gpt-5-mini",
+                    "output": [],
+                    "usage": {
+                        "input_tokens": 100,
+                        "output_tokens": 50,
+                        "total_tokens": 150,
+                        "cost": 0.01234,
+                    },
+                },
+            },
+            logging_obj=Mock(),
+        )
+
+        assert (
+            result.response._hidden_params["additional_headers"]["llm_provider-x-litellm-response-cost"]
+            == 0.01234
+        )
+
 
 class TestOpenRouterResponsesAPIRegistration:
     """Test that OpenRouter is properly registered as a native Responses API provider."""