fix: fix responses cost calc

litellm/litellm_core_utils/litellm_logging.py

@@ -1455,6 +1455,20 @@ class Logging(LiteLLMLoggingBaseClass):
             ):  # use model_id if not already set
                 router_model_id = hidden_params["model_id"]
 
+        # Fallback: extract router_model_id from litellm_params when not available
+        # from the result object. ResponsesAPIResponse objects (used by /v1/responses
+        # streaming) don't carry _hidden_params["model_id"] like ModelResponse does.
+        if router_model_id is None and hasattr(self, "litellm_params"):
+            for metadata_key in ("litellm_metadata", "metadata"):
+                _metadata: dict = (
+                    self.litellm_params.get(metadata_key, {}) or {}
+                )
+                _model_info: dict = _metadata.get("model_info", {}) or {}
+                _model_id = _model_info.get("id")
+                if _model_id is not None:
+                    router_model_id = _model_id
+                    break
+
         ## RESPONSE COST ##
         custom_pricing = use_custom_pricing_for_model(
             litellm_params=(

litellm/proxy/_new_secret_config.yaml

@@ -1,41 +1,32 @@
 model_list:
-  - model_name: gpt-3.5-turbo
-    litellm_params:
-      model: openai/gpt-3.5-turbo
-      api_key: os.environ/OPENAI_API_KEY
-  - model_name: gpt-4o
-    litellm_params:
-      model: openai/gpt-4o
-      api_key: os.environ/OPENAI_API_KEY
-  - model_name: claude-sonnet-4-5-20250929
-    litellm_params:
-      model: anthropic/claude-sonnet-4-5-20250929
-  - model_name: gpt-4.1-mini
+
+  # OpenAI model for /v1/chat/completions test — 200x custom pricing
+  - model_name: "gpt-4.1-mini"
     litellm_params:
       model: openai/gpt-4.1-mini
-  - model_name: gpt-5-mini
+      api_key: os.environ/OPENAI_API_KEY
+    model_info:
+      id: gpt-4.1-mini-custom-pricing
+      input_cost_per_token: 0.00004        # 100x standard ($0.40/1M = $0.0000004)
+      output_cost_per_token: 0.00016       # 100x standard ($1.60/1M = $0.0000016)
+
+  # OpenAI model for /v1/responses test — 100x custom pricing
+  - model_name: "gpt-5"
     litellm_params:
-      model: openai/gpt-5-mini
-  - model_name: custom_litellm_model
+      model: openai/gpt-5
+      api_key: os.environ/OPENAI_API_KEY
+    model_info:
+      id: gpt-5-custom-pricing
+      mode: "chat"
+      input_cost_per_token: 125        # 100x standard ($1.25/1M = $0.00000125)
+      output_cost_per_token: 10         # 100x standard ($10.00/1M = $0.00001)
+
+  # Anthropic model for /v1/messages test — 100x custom pricing
+  - model_name: "claude-sonnet-4-20250514"
     litellm_params:
-      model: litellm_agent/claude-sonnet-4-5-20250929
-      litellm_system_prompt: "Be a helpful assistant."
-
-
-guardrails:
-  - guardrail_name: "tool_policy"
-    litellm_params:
-      guardrail: tool_policy
-      mode: [pre_call, post_call]
-      default_on: true
-
-mcp_servers:
-  my_http_server:
-    url: "http://0.0.0.0:8001/mcp"
-    transport: "http"
-    description: "My custom MCP server"
-    available_on_public_internet: true
-
-general_settings:
-  store_model_in_db: true
-  store_prompts_in_spend_logs: true
+      model: anthropic/claude-sonnet-4-20250514
+      api_key: os.environ/ANTHROPIC_API_KEY
+    model_info:
+      id: claude-sonnet-4-custom-pricing
+      input_cost_per_token: 0.0003         # 100x standard ($0.000003)
+      output_cost_per_token: 0.0015         # 100x standard ($0.000015)

tests/test_litellm/litellm_core_utils/test_litellm_logging.py

@@ -180,6 +180,82 @@ def test_use_custom_pricing_not_detected_litellm_metadata_no_pricing():
     assert use_custom_pricing_for_model(litellm_params) is False
 
 
+def test_response_cost_calculator_uses_router_model_id_from_litellm_metadata():
+    """_response_cost_calculator should extract router_model_id from
+    litellm_params.litellm_metadata.model_info.id when the result object
+    does not carry _hidden_params (e.g. ResponsesAPIResponse from /v1/responses
+    streaming). Regression test for custom pricing on streaming responses."""
+    import litellm
+    from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
+    from litellm.types.llms.openai import ResponsesAPIResponse
+
+    custom_model_id = "gpt-5-custom-pricing"
+    custom_input_cost = 125.0
+    custom_output_cost = 10.0
+
+    litellm.register_model(
+        model_cost={
+            custom_model_id: {
+                "input_cost_per_token": custom_input_cost,
+                "output_cost_per_token": custom_output_cost,
+                "max_tokens": 128000,
+                "max_input_tokens": 128000,
+                "max_output_tokens": 16384,
+                "litellm_provider": "openai",
+            }
+        }
+    )
+
+    try:
+        logging_obj = LiteLLMLoggingObj(
+            model="gpt-5",
+            messages=[{"role": "user", "content": "Hi"}],
+            stream=True,
+            call_type="aresponses",
+            start_time=time.time(),
+            litellm_call_id="test-123",
+            function_id="test-fn",
+        )
+
+        logging_obj.update_environment_variables(
+            model="gpt-5",
+            user="",
+            optional_params={},
+            litellm_params={
+                "api_base": "",
+                "litellm_metadata": {
+                    "model_info": {
+                        "id": custom_model_id,
+                        "input_cost_per_token": custom_input_cost,
+                        "output_cost_per_token": custom_output_cost,
+                    },
+                },
+            },
+        )
+
+        response_obj = ResponsesAPIResponse(
+            id="resp_abc",
+            created_at=1234567890,
+            model="gpt-5",
+            output=[],
+            usage={
+                "input_tokens": 10,
+                "output_tokens": 5,
+                "total_tokens": 15,
+            },
+        )
+
+        cost = logging_obj._response_cost_calculator(result=response_obj)
+
+        assert cost is not None, "Cost should not be None"
+        expected_cost = (10 * custom_input_cost) + (5 * custom_output_cost)
+        assert cost == pytest.approx(
+            expected_cost
+        ), f"Expected {expected_cost}, got {cost}"
+    finally:
+        litellm.model_cost.pop(custom_model_id, None)
+
+
 class TestUpdateFromKwargs:
     """Tests for the update_from_kwargs convenience wrapper."""