diff --git a/litellm/proxy/pass_through_endpoints/llm_provider_handlers/openai_websocket_passthrough_logging_handler.py b/litellm/proxy/pass_through_endpoints/llm_provider_handlers/openai_websocket_passthrough_logging_handler.py
index c7df3259ce2..d3e145f8751 100644
--- a/litellm/proxy/pass_through_endpoints/llm_provider_handlers/openai_websocket_passthrough_logging_handler.py
+++ b/litellm/proxy/pass_through_endpoints/llm_provider_handlers/openai_websocket_passthrough_logging_handler.py
@@ -23,6 +23,7 @@ _SESSION_CREATED: Final = "session.created"
 _REALTIME_RESPONSE_DONE: Final = "response.done"
 _RESPONSES_COMPLETED: Final = "response.completed"
 _MODEL_NAMING_EVENTS: Final = frozenset({_SESSION_CREATED, _REALTIME_RESPONSE_DONE, _RESPONSES_COMPLETED})
+_RESPONSE_COST_HEADER: Final = "llm_provider-x-litellm-response-cost"
 
 
 class _NamedModel(BaseModel):
@@ -148,6 +149,21 @@ def _responses_cost(usage: Usage, model: str) -> float:
         return 0.0
 
 
+def _priced_result(model: str, billing: _Billing, start_time: datetime) -> ModelResponse:
+    result: Final = ModelResponse(
+        id=f"openai-websocket-{start_time.timestamp()}",
+        object="chat.completion",
+        created=int(start_time.timestamp()),
+        model=model,
+        usage=billing.usage,
+    )
+    hidden: Final = {  # mutable-ok: response_cost_calculator writes optional_params into this dict
+        "additional_headers": MappingProxyType({_RESPONSE_COST_HEADER: billing.cost})
+    }
+    result._hidden_params = hidden  # pyright: ignore[reportPrivateUsage]  # the one settled-cost channel
+    return result
+
+
 class OpenAIWebsocketPassthroughLoggingHandler:
     def openai_websocket_passthrough_handler(
         self,
@@ -181,13 +197,7 @@ class OpenAIWebsocketPassthroughLoggingHandler:
 
         logging_obj.model_call_details.update(MappingProxyType({"response_cost": billing.cost}))
         priced: Final[PassThroughEndpointLoggingTypedDict] = {
-            "result": ModelResponse(
-                id=f"openai-websocket-{start_time.timestamp()}",
-                object="chat.completion",
-                created=int(start_time.timestamp()),
-                model=model,
-                usage=billing.usage,
-            ),
+            "result": _priced_result(model, billing, start_time),
             "kwargs": {
                 **kwargs,
                 "model": model,
diff --git a/tests/test_litellm/proxy/pass_through_endpoints/llm_provider_handlers/test_openai_websocket_passthrough_logging_handler.py b/tests/test_litellm/proxy/pass_through_endpoints/llm_provider_handlers/test_openai_websocket_passthrough_logging_handler.py
index ca0268950f7..7aeb68df26d 100644
--- a/tests/test_litellm/proxy/pass_through_endpoints/llm_provider_handlers/test_openai_websocket_passthrough_logging_handler.py
+++ b/tests/test_litellm/proxy/pass_through_endpoints/llm_provider_handlers/test_openai_websocket_passthrough_logging_handler.py
@@ -237,3 +237,33 @@ def test_a_responses_turn_without_readable_usage_is_not_billed():
 
     assert handled["result"] is None
     assert handled["kwargs"] == {"model": RESPONSES_MODEL, "custom_llm_provider": "openai"}
+
+
+def test_the_price_of_a_multi_model_connection_reaches_the_spend_row():
+    """
+    The logging layer prices whatever result it is handed, and one result can
+    only name one model, so a connection that ran two models has to carry the
+    price it already worked out or the cheaper model pays for both turns.
+    """
+    handled = _handle(
+        [
+            _responses_completed(RESPONSES_MODEL, input_tokens=13, output_tokens=5),
+            _responses_completed(REASONING_MODEL, input_tokens=20, output_tokens=7),
+        ],
+        url_route="/openai_passthrough/v1/responses",
+    )
+
+    logged_cost = litellm.response_cost_calculator(
+        response_object=handled["result"],
+        model=handled["kwargs"]["model"],
+        custom_llm_provider="openai",
+        call_type="completion",
+        optional_params={},
+    )
+
+    session_prices = litellm.model_cost[RESPONSES_MODEL]
+    priced_on_the_session_model = (
+        33 * session_prices["input_cost_per_token"] + 12 * session_prices["output_cost_per_token"]
+    )
+    assert logged_cost == pytest.approx(handled["kwargs"]["response_cost"])
+    assert logged_cost > priced_on_the_session_model