From c958f9db7e0a024f7fbbb28168c47ac9e65c4853 Mon Sep 17 00:00:00 2001
From: kerry <kerry@berri.ai>
Date: Sat, 19 Sep 2026 18:44:51 +0000
Subject: [PATCH 1/6] test: extend cost tracking integration harness

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---
 tests/integration/_support/upstream.py        |  20 +--
 tests/integration/contracts.json              |   9 ++
 .../integration/cost_calculation/conftest.py  |  34 ++++-
 .../cost_calculation/cost_tracking_case.py    |  74 +++++++++-
 .../cost_calculation/cost_tracking_cases.json | 135 +++++++++++++++++-
 .../cost_calculation/test_cost_tracking.py    |  54 ++++++-
 6 files changed, 304 insertions(+), 22 deletions(-)

diff --git a/tests/integration/_support/upstream.py b/tests/integration/_support/upstream.py
index 1ad02b6a3f2..3c4198c8133 100644
--- a/tests/integration/_support/upstream.py
+++ b/tests/integration/_support/upstream.py
@@ -1,25 +1,19 @@
 from __future__ import annotations
 
 import argparse
+import json
+import os
+import struct
+import zlib
 from collections import deque
 from collections.abc import Mapping
-import json
 from dataclasses import dataclass, field
-import os
 from pathlib import Path
 from queue import SimpleQueue
-import struct
 from typing import Final, cast
-import zlib
 
 import httpx
 import uvicorn
-from pydantic import BaseModel, JsonValue, TypeAdapter, ValidationError
-from starlette.applications import Starlette
-from starlette.requests import Request
-from starlette.responses import JSONResponse, Response
-from starlette.routing import Route
-
 from _fake_openai_endpoint_server import chat_completions, completions, embeddings, health, moderations
 from integration.cost_calculation.cost_tracking_case import (
     EventStreamResponse,
@@ -27,6 +21,11 @@ from integration.cost_calculation.cost_tracking_case import (
     SseResponse,
     StoredResponse,
 )
+from pydantic import BaseModel, JsonValue, TypeAdapter, ValidationError
+from starlette.applications import Starlette
+from starlette.requests import Request
+from starlette.responses import JSONResponse, Response
+from starlette.routing import Route
 
 JSON_OBJECT: Final = TypeAdapter(dict[str, JsonValue])
 CASES_FILE: Final = Path(__file__).resolve().parents[1] / "cost_calculation" / "cost_tracking_cases.json"
@@ -210,6 +209,7 @@ class Provider:
                         "$REQUEST_ID", scenario_id
                     ).encode(),
                     media_type=response.content_type,
+                    status_code=response.status,
                 )
             case SseResponse():
                 stream_body: Final = ("\n\n".join(response.frames) + "\n\n").replace(
diff --git a/tests/integration/contracts.json b/tests/integration/contracts.json
index fe7b6dfe7ac..1b0c9e31d4e 100644
--- a/tests/integration/contracts.json
+++ b/tests/integration/contracts.json
@@ -382,6 +382,15 @@
     "tests/integration/cost_calculation/test_cost_tracking.py::test_case_bills_expected_cost[azure-gpt-5.6-stream_full_usage]": [
       "quota_management.spend_tracking.scripted_wire.logs_cost"
     ],
+    "tests/integration/cost_calculation/test_cost_tracking.py::test_case_bills_expected_cost[gpt-5.6-responses_native_json]": [
+      "quota_management.spend_tracking.cost_matrix.logs_cost"
+    ],
+    "tests/integration/cost_calculation/test_cost_tracking.py::test_case_bills_expected_cost[gpt-5.6-upstream_500_zero_spend]": [
+      "quota_management.spend_tracking.cost_matrix.logs_cost"
+    ],
+    "tests/integration/cost_calculation/test_cost_tracking.py::test_case_bills_expected_cost[gpt-5.6-upstream_429_zero_spend]": [
+      "quota_management.spend_tracking.cost_matrix.logs_cost"
+    ],
     "tests/integration/cost_calculation/test_cost_tracking.py::test_case_bills_expected_cost[claude-haiku-4-5-input_text]": [
       "quota_management.spend_tracking.cost_matrix.logs_cost"
     ],
diff --git a/tests/integration/cost_calculation/conftest.py b/tests/integration/cost_calculation/conftest.py
index f1b8901d626..192f2bf5461 100644
--- a/tests/integration/cost_calculation/conftest.py
+++ b/tests/integration/cost_calculation/conftest.py
@@ -9,12 +9,11 @@ from typing import Final
 
 from cryptography.hazmat.primitives import serialization
 from cryptography.hazmat.primitives.asymmetric import rsa
-from pydantic import BaseModel, ConfigDict
-
 from integration._support.client import JSON_OBJECT, Scenario, eventually, object_value, string_value
 from integration._support.database import read_rows
 from integration._support.upstream import delete_scenario, register_scenario
 from integration.cost_calculation.cost_tracking_case import CostTrackingTestCase
+from pydantic import BaseModel, ConfigDict
 
 
 class CostBreakdown(BaseModel):
@@ -50,6 +49,15 @@ class CostRow(BaseModel):
         return self.metadata.cost_breakdown
 
 
+class FailureRow(BaseModel):
+    model_config = ConfigDict(extra="ignore")
+
+    spend: float
+    status: str
+    prompt_tokens: int | None = None
+    completion_tokens: int | None = None
+
+
 def approx_equal(actual: float, expected: float) -> bool:
     return abs(actual - expected) <= max(1e-9, abs(expected) * 1e-2)
 
@@ -92,6 +100,28 @@ def poll_cost_row(key: str) -> CostRow:
     return result
 
 
+def poll_failure_row(key: str) -> FailureRow:
+    digest: Final = sha256(key.encode()).hexdigest()
+
+    def read() -> FailureRow | None:
+        rows: Final = read_rows(
+            'SELECT spend, status, prompt_tokens, completion_tokens FROM "LiteLLM_SpendLogs" WHERE api_key=%s',
+            (digest,),
+        )
+        return next(
+            (
+                parsed
+                for row in rows
+                if (parsed := FailureRow.model_validate(row)).status == "failure"
+            ),
+            None,
+        )
+
+    result: Final = eventually(read, lambda row: row is not None, seconds=60)
+    assert result is not None
+    return result
+
+
 @functools.cache
 def _vertex_private_key_pem() -> str:
     return rsa.generate_private_key(public_exponent=65537, key_size=2048).private_bytes(
diff --git a/tests/integration/cost_calculation/cost_tracking_case.py b/tests/integration/cost_calculation/cost_tracking_case.py
index 6af95f995ff..cb408f2c488 100644
--- a/tests/integration/cost_calculation/cost_tracking_case.py
+++ b/tests/integration/cost_calculation/cost_tracking_case.py
@@ -71,6 +71,7 @@ class JsonResponse(BaseModel):
 
     content_type: Literal["application/json"]
     body: dict[str, JsonValue]
+    status: int = 200
 
 
 class SseResponse(BaseModel):
@@ -108,6 +109,10 @@ class ExactExpected(BaseModel):
     output_cost: float
     prompt_tokens: int
     completion_tokens: int
+    cache_read_cost: float | None = None
+    cache_creation_cost: float | None = None
+    reasoning_cost: float | None = None
+    tool_usage_cost: float | None = None
 
 
 class RecountRates(BaseModel):
@@ -123,7 +128,19 @@ class RecountExpected(BaseModel):
     recount: RecountRates
 
 
-Expected: TypeAlias = ExactExpected | RecountExpected
+class FailureDetails(BaseModel):
+    model_config = ConfigDict(frozen=True, extra="forbid")
+
+    status: int
+
+
+class FailureExpected(BaseModel):
+    model_config = ConfigDict(frozen=True, extra="forbid")
+
+    failure: FailureDetails
+
+
+Expected: TypeAlias = ExactExpected | RecountExpected | FailureExpected
 
 
 class CostTrackingTestCase(BaseModel):
@@ -132,6 +149,15 @@ class CostTrackingTestCase(BaseModel):
     name: str
     covers: str
     model: str
+    endpoint: Literal[
+        "/v1/chat/completions",
+        "/v1/responses",
+        "/v1/messages",
+        "/v1/embeddings",
+        "/v1/rerank",
+        "/v1/completions",
+        "/v1/moderations",
+    ] = "/v1/chat/completions"
     deployment: Deployment | None = None
     request: dict[str, JsonValue]
     response: StoredResponse
@@ -240,6 +266,44 @@ def data_errors() -> tuple[str, ...]:
             or case.expected.recount.output_cost_per_token != (COST_MAP[case.model].output_cost_per_token or 0.0)
         )
     )
+    component_mismatches: Final = sorted(
+        case.name
+        for case in CASES
+        if isinstance(case.expected, ExactExpected)
+        and any(
+            component is not None
+            for component in (
+                case.expected.cache_read_cost,
+                case.expected.cache_creation_cost,
+                case.expected.reasoning_cost,
+                case.expected.tool_usage_cost,
+            )
+        )
+        and (
+            (case.expected.cache_read_cost or 0.0) + (case.expected.cache_creation_cost or 0.0)
+            > case.expected.input_cost
+            or (case.expected.reasoning_cost or 0.0) > case.expected.output_cost
+            or not _approx_equal(
+                case.expected.input_cost
+                + case.expected.output_cost
+                + (case.expected.tool_usage_cost or 0.0),
+                case.expected.spend,
+            )
+        )
+    )
+    failure_response_mismatches: Final = sorted(
+        case.name
+        for case in CASES
+        if (
+            isinstance(case.expected, FailureExpected)
+            and (not isinstance(case.response, JsonResponse) or case.response.status < 400)
+        )
+        or (
+            not isinstance(case.expected, FailureExpected)
+            and isinstance(case.response, JsonResponse)
+            and case.response.status != 200
+        )
+    )
     return tuple(
         message
         for message in (
@@ -248,6 +312,14 @@ def data_errors() -> tuple[str, ...]:
             f"duplicate case names: {duplicate_names}" if duplicate_names else None,
             f"cost-map entries share input_cost_per_token: {shared_input_rates}" if shared_input_rates else None,
             f"recount rates differ from cost-map rates: {recount_mismatches}" if recount_mismatches else None,
+            f"breakdown components are inconsistent: {component_mismatches}" if component_mismatches else None,
+            f"failure response statuses are inconsistent: {failure_response_mismatches}"
+            if failure_response_mismatches
+            else None,
         )
         if message is not None
     )
+
+
+def _approx_equal(actual: float, expected: float) -> bool:
+    return abs(actual - expected) <= max(1e-9, abs(expected) * 1e-2)
diff --git a/tests/integration/cost_calculation/cost_tracking_cases.json b/tests/integration/cost_calculation/cost_tracking_cases.json
index 3627774816f..379250d3123 100644
--- a/tests/integration/cost_calculation/cost_tracking_cases.json
+++ b/tests/integration/cost_calculation/cost_tracking_cases.json
@@ -534,7 +534,8 @@
         "input_cost": 0.00616704,
         "output_cost": 0.00627,
         "prompt_tokens": 12928,
-        "completion_tokens": 380
+        "completion_tokens": 380,
+        "cache_read_cost": 0.00405504
       }
     },
     {
@@ -605,7 +606,8 @@
         "input_cost": 0.0397056,
         "output_cost": 0.005775,
         "prompt_tokens": 9728,
-        "completion_tokens": 350
+        "completion_tokens": 350,
+        "cache_creation_cost": 0.038016
       }
     },
     {
@@ -681,7 +683,8 @@
         "input_cost": 0.0574464,
         "output_cost": 0.005775,
         "prompt_tokens": 9728,
-        "completion_tokens": 350
+        "completion_tokens": 350,
+        "cache_creation_cost": 0.0557568
       }
     },
     {
@@ -3417,7 +3420,8 @@
         "input_cost": 0.002232,
         "output_cost": 0.065484,
         "prompt_tokens": 1240,
-        "completion_tokens": 4040
+        "completion_tokens": 4040,
+        "reasoning_cost": 0.05742
       }
     },
     {
@@ -3638,7 +3642,8 @@
         "input_cost": 0.003312,
         "output_cost": 0.0059328,
         "prompt_tokens": 1840,
-        "completion_tokens": 412
+        "completion_tokens": 412,
+        "tool_usage_cost": 0.0125
       }
     },
     {
@@ -4494,7 +4499,8 @@
         "input_cost": 0.0018688,
         "output_cost": 0.0019,
         "prompt_tokens": 12928,
-        "completion_tokens": 380
+        "completion_tokens": 380,
+        "cache_read_cost": 0.0012288
       }
     },
     {
@@ -16955,7 +16961,8 @@
         "input_cost": 0.00276,
         "output_cost": 0.004944,
         "prompt_tokens": 1840,
-        "completion_tokens": 412
+        "completion_tokens": 412,
+        "tool_usage_cost": 0.0025
       }
     },
     {
@@ -21797,6 +21804,120 @@
         "completion_tokens": 1592
       }
     },
+    {
+      "name": "gpt-5.6-responses_native_json",
+      "covers": "quota_management.spend_tracking.cost_matrix.logs_cost",
+      "model": "gpt-5.6",
+      "endpoint": "/v1/responses",
+      "request": {
+        "model": "$MODEL",
+        "input": "responses native fixture",
+        "stream": false
+      },
+      "response": {
+        "content_type": "application/json",
+        "body": {
+          "object": "response",
+          "status": "completed",
+          "model": "gpt-5.6",
+          "output": [
+            {
+              "type": "message",
+              "id": "msg_$REQUEST_ID",
+              "status": "completed",
+              "role": "assistant",
+              "content": [
+                {
+                  "type": "output_text",
+                  "text": "scripted response",
+                  "annotations": []
+                }
+              ]
+            }
+          ],
+          "usage": {
+            "input_tokens": 11,
+            "output_tokens": 7,
+            "total_tokens": 18,
+            "input_tokens_details": {
+              "cached_tokens": 0
+            },
+            "output_tokens_details": {
+              "reasoning_tokens": 0
+            }
+          }
+        }
+      },
+      "expected": {
+        "spend": 0.00011725,
+        "input_cost": 1.925e-05,
+        "output_cost": 9.8e-05,
+        "prompt_tokens": 11,
+        "completion_tokens": 7
+      }
+    },
+    {
+      "name": "gpt-5.6-upstream_500_zero_spend",
+      "covers": "quota_management.spend_tracking.cost_matrix.logs_cost",
+      "model": "gpt-5.6",
+      "request": {
+        "model": "$MODEL",
+        "messages": [
+          {
+            "role": "user",
+            "content": "scripted upstream failure 500"
+          }
+        ],
+        "stream": false
+      },
+      "response": {
+        "content_type": "application/json",
+        "status": 500,
+        "body": {
+          "error": {
+            "message": "scripted upstream failure",
+            "type": "server_error",
+            "code": "500"
+          }
+        }
+      },
+      "expected": {
+        "failure": {
+          "status": 500
+        }
+      }
+    },
+    {
+      "name": "gpt-5.6-upstream_429_zero_spend",
+      "covers": "quota_management.spend_tracking.cost_matrix.logs_cost",
+      "model": "gpt-5.6",
+      "request": {
+        "model": "$MODEL",
+        "messages": [
+          {
+            "role": "user",
+            "content": "scripted upstream failure 429"
+          }
+        ],
+        "stream": false
+      },
+      "response": {
+        "content_type": "application/json",
+        "status": 429,
+        "body": {
+          "error": {
+            "message": "scripted upstream failure",
+            "type": "rate_limit_error",
+            "code": "429"
+          }
+        }
+      },
+      "expected": {
+        "failure": {
+          "status": 429
+        }
+      }
+    },
     {
       "name": "meta.llama4-maverick-17b-instruct-v1:0-input_text",
       "covers": "quota_management.spend_tracking.cost_matrix.logs_cost",
diff --git a/tests/integration/cost_calculation/test_cost_tracking.py b/tests/integration/cost_calculation/test_cost_tracking.py
index a8a56fbfbbd..c9dc7a5ffe3 100644
--- a/tests/integration/cost_calculation/test_cost_tracking.py
+++ b/tests/integration/cost_calculation/test_cost_tracking.py
@@ -6,18 +6,19 @@ from hashlib import sha256
 from typing import Final, cast
 
 import pytest
-
 from integration._support.client import JSON_OBJECT, Gateway
 from integration.cost_calculation.conftest import (
     approx_equal,
     assert_total_is_sum_of_components,
     poll_cost_row,
+    poll_failure_row,
     register_scenario_deployment,
 )
 from integration.cost_calculation.cost_tracking_case import (
     CASES,
     CostTrackingTestCase,
     ExactExpected,
+    FailureExpected,
     RecountExpected,
     data_errors,
 )
@@ -51,10 +52,22 @@ def test_case_bills_expected_cost(gateway: Gateway, case: CostTrackingTestCase)
         model_name: Final = register_scenario_deployment(scenario, case, marker, key)
         response: Final = gateway.request(
             "POST",
-            "/v1/chat/completions",
+            case.endpoint,
             {**case.request, "model": model_name},
             key=key,
         )
+        if isinstance(case.expected, FailureExpected):
+            assert response.status_code == case.expected.failure.status, (
+                f"{case.name}: proxy returned {response.status_code}, expected {case.expected.failure.status}: "
+                f"{response.text[:400]}"
+            )
+            response_cost: Final = response.headers.get("x-litellm-response-cost")
+            assert response_cost is None or approx_equal(float(response_cost), 0.0), (
+                f"{case.name}: failure response cost was {response_cost}"
+            )
+            row: Final = poll_failure_row(key)
+            assert row.spend == 0, f"{case.name}: failure spend was {row.spend}"
+            return
         assert response.is_success, f"{case.name}: proxy returned {response.status_code}: {response.text[:400]}"
         if case.response.content_type == "text/event-stream":
             _assert_stream_has_no_error(response.text)
@@ -92,6 +105,43 @@ def test_case_bills_expected_cost(gateway: Gateway, case: CostTrackingTestCase)
         assert breakdown.output_cost is not None and approx_equal(breakdown.output_cost, expected.output_cost), (
             f"{case.name}: output_cost {breakdown.output_cost} != expected {expected.output_cost}"
         )
+        for field, header_name, expected_component in (
+            ("cache_read_cost", "x-litellm-response-cost-cache-read", expected.cache_read_cost),
+            ("cache_creation_cost", "x-litellm-response-cost-cache-creation", expected.cache_creation_cost),
+            ("reasoning_cost", "x-litellm-response-cost-reasoning", expected.reasoning_cost),
+            ("tool_usage_cost", "x-litellm-response-cost-tool-usage", expected.tool_usage_cost),
+        ):
+            if expected_component is None:
+                continue
+            actual_component: Final = getattr(breakdown, field)
+            assert actual_component is not None and approx_equal(actual_component, expected_component), (
+                f"{case.name}: {field} {actual_component} != expected {expected_component}"
+            )
+            if case.response.content_type == "application/json":
+                header: Final = response.headers.get(header_name)
+                assert header is not None and approx_equal(float(header), expected_component), (
+                    f"{case.name}: {header_name} {header} != expected {expected_component}"
+                )
+        if case.response.content_type == "application/json" and any(
+            component is not None
+            for component in (
+                expected.cache_read_cost,
+                expected.cache_creation_cost,
+                expected.reasoning_cost,
+                expected.tool_usage_cost,
+            )
+        ):
+            input_header: Final = response.headers.get("x-litellm-response-cost-input")
+            output_header: Final = response.headers.get("x-litellm-response-cost-output")
+            expected_input_header: Final = expected.input_cost - (
+                expected.cache_read_cost or 0.0
+            ) - (expected.cache_creation_cost or 0.0)
+            assert input_header is not None and approx_equal(float(input_header), expected_input_header), (
+                f"{case.name}: x-litellm-response-cost-input {input_header} != expected {expected_input_header}"
+            )
+            assert output_header is not None and approx_equal(float(output_header), expected.output_cost), (
+                f"{case.name}: x-litellm-response-cost-output {output_header} != expected {expected.output_cost}"
+            )
         assert row.prompt_tokens == expected.prompt_tokens, (
             f"{case.name}: prompt_tokens {row.prompt_tokens} != expected {expected.prompt_tokens}"
         )

From e42f3f1562c9b5a07ef69a1468269399beb4bc24 Mon Sep 17 00:00:00 2001
From: kerry <kerry@berri.ai>
Date: Sat, 19 Sep 2026 19:10:04 +0000
Subject: [PATCH 2/6] test: fix cost harness review issues

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---
 tests/integration/_support/upstream.py        | 19 +++++------
 .../integration/cost_calculation/conftest.py  |  3 +-
 .../cost_calculation/cost_tracking_cases.json |  2 ++
 .../cost_calculation/test_cost_tracking.py    | 32 +++++++++++++++----
 4 files changed, 40 insertions(+), 16 deletions(-)

diff --git a/tests/integration/_support/upstream.py b/tests/integration/_support/upstream.py
index 3c4198c8133..a289589b2dc 100644
--- a/tests/integration/_support/upstream.py
+++ b/tests/integration/_support/upstream.py
@@ -1,19 +1,25 @@
 from __future__ import annotations
 
 import argparse
-import json
-import os
-import struct
-import zlib
 from collections import deque
 from collections.abc import Mapping
+import json
 from dataclasses import dataclass, field
+import os
 from pathlib import Path
 from queue import SimpleQueue
+import struct
 from typing import Final, cast
+import zlib
 
 import httpx
 import uvicorn
+from pydantic import BaseModel, JsonValue, TypeAdapter, ValidationError
+from starlette.applications import Starlette
+from starlette.requests import Request
+from starlette.responses import JSONResponse, Response
+from starlette.routing import Route
+
 from _fake_openai_endpoint_server import chat_completions, completions, embeddings, health, moderations
 from integration.cost_calculation.cost_tracking_case import (
     EventStreamResponse,
@@ -21,11 +27,6 @@ from integration.cost_calculation.cost_tracking_case import (
     SseResponse,
     StoredResponse,
 )
-from pydantic import BaseModel, JsonValue, TypeAdapter, ValidationError
-from starlette.applications import Starlette
-from starlette.requests import Request
-from starlette.responses import JSONResponse, Response
-from starlette.routing import Route
 
 JSON_OBJECT: Final = TypeAdapter(dict[str, JsonValue])
 CASES_FILE: Final = Path(__file__).resolve().parents[1] / "cost_calculation" / "cost_tracking_cases.json"
diff --git a/tests/integration/cost_calculation/conftest.py b/tests/integration/cost_calculation/conftest.py
index 192f2bf5461..166488e36a5 100644
--- a/tests/integration/cost_calculation/conftest.py
+++ b/tests/integration/cost_calculation/conftest.py
@@ -9,11 +9,12 @@ from typing import Final
 
 from cryptography.hazmat.primitives import serialization
 from cryptography.hazmat.primitives.asymmetric import rsa
+from pydantic import BaseModel, ConfigDict
+
 from integration._support.client import JSON_OBJECT, Scenario, eventually, object_value, string_value
 from integration._support.database import read_rows
 from integration._support.upstream import delete_scenario, register_scenario
 from integration.cost_calculation.cost_tracking_case import CostTrackingTestCase
-from pydantic import BaseModel, ConfigDict
 
 
 class CostBreakdown(BaseModel):
diff --git a/tests/integration/cost_calculation/cost_tracking_cases.json b/tests/integration/cost_calculation/cost_tracking_cases.json
index 379250d3123..428124ac4a9 100644
--- a/tests/integration/cost_calculation/cost_tracking_cases.json
+++ b/tests/integration/cost_calculation/cost_tracking_cases.json
@@ -21817,8 +21817,10 @@
       "response": {
         "content_type": "application/json",
         "body": {
+          "id": "resp_$REQUEST_ID",
           "object": "response",
           "status": "completed",
+          "created_at": 1700000000,
           "model": "gpt-5.6",
           "output": [
             {
diff --git a/tests/integration/cost_calculation/test_cost_tracking.py b/tests/integration/cost_calculation/test_cost_tracking.py
index c9dc7a5ffe3..5b876974c63 100644
--- a/tests/integration/cost_calculation/test_cost_tracking.py
+++ b/tests/integration/cost_calculation/test_cost_tracking.py
@@ -6,6 +6,7 @@ from hashlib import sha256
 from typing import Final, cast
 
 import pytest
+
 from integration._support.client import JSON_OBJECT, Gateway
 from integration.cost_calculation.conftest import (
     approx_equal,
@@ -105,15 +106,34 @@ def test_case_bills_expected_cost(gateway: Gateway, case: CostTrackingTestCase)
         assert breakdown.output_cost is not None and approx_equal(breakdown.output_cost, expected.output_cost), (
             f"{case.name}: output_cost {breakdown.output_cost} != expected {expected.output_cost}"
         )
-        for field, header_name, expected_component in (
-            ("cache_read_cost", "x-litellm-response-cost-cache-read", expected.cache_read_cost),
-            ("cache_creation_cost", "x-litellm-response-cost-cache-creation", expected.cache_creation_cost),
-            ("reasoning_cost", "x-litellm-response-cost-reasoning", expected.reasoning_cost),
-            ("tool_usage_cost", "x-litellm-response-cost-tool-usage", expected.tool_usage_cost),
+        for field, header_name, actual_component, expected_component in (
+            (
+                "cache_read_cost",
+                "x-litellm-response-cost-cache-read",
+                breakdown.cache_read_cost,
+                expected.cache_read_cost,
+            ),
+            (
+                "cache_creation_cost",
+                "x-litellm-response-cost-cache-creation",
+                breakdown.cache_creation_cost,
+                expected.cache_creation_cost,
+            ),
+            (
+                "reasoning_cost",
+                "x-litellm-response-cost-reasoning",
+                breakdown.reasoning_cost,
+                expected.reasoning_cost,
+            ),
+            (
+                "tool_usage_cost",
+                "x-litellm-response-cost-tool-usage",
+                breakdown.tool_usage_cost,
+                expected.tool_usage_cost,
+            ),
         ):
             if expected_component is None:
                 continue
-            actual_component: Final = getattr(breakdown, field)
             assert actual_component is not None and approx_equal(actual_component, expected_component), (
                 f"{case.name}: {field} {actual_component} != expected {expected_component}"
             )

From dccb1b56b47c62fdca4865c967d7901a0cd2a461 Mon Sep 17 00:00:00 2001
From: kerry <kerry@berri.ai>
Date: Sat, 19 Sep 2026 19:17:46 +0000
Subject: [PATCH 3/6] test(integration): reject out-of-range failure statuses
 at import

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---
 tests/integration/cost_calculation/cost_tracking_case.py | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/tests/integration/cost_calculation/cost_tracking_case.py b/tests/integration/cost_calculation/cost_tracking_case.py
index cb408f2c488..9737508e27c 100644
--- a/tests/integration/cost_calculation/cost_tracking_case.py
+++ b/tests/integration/cost_calculation/cost_tracking_case.py
@@ -296,7 +296,11 @@ def data_errors() -> tuple[str, ...]:
         for case in CASES
         if (
             isinstance(case.expected, FailureExpected)
-            and (not isinstance(case.response, JsonResponse) or case.response.status < 400)
+            and (
+                not isinstance(case.response, JsonResponse)
+                or not 400 <= case.response.status <= 599
+                or not 400 <= case.expected.failure.status <= 599
+            )
         )
         or (
             not isinstance(case.expected, FailureExpected)

From 9bd648baf590ccf5ad3f81ccc44e29a57c3bc7b9 Mon Sep 17 00:00:00 2001
From: kerry <kerry@berri.ai>
Date: Sat, 19 Sep 2026 19:22:25 +0000
Subject: [PATCH 4/6] test(integration): price fireworks cached input at the
 50% default

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---
 tests/integration/contracts.json                           | 2 +-
 .../integration/cost_calculation/cost_tracking_cases.json  | 7 ++++---
 2 files changed, 5 insertions(+), 4 deletions(-)

diff --git a/tests/integration/contracts.json b/tests/integration/contracts.json
index 1b0c9e31d4e..338ace07908 100644
--- a/tests/integration/contracts.json
+++ b/tests/integration/contracts.json
@@ -559,7 +559,7 @@
     "tests/integration/cost_calculation/test_cost_tracking.py::test_case_bills_expected_cost[fireworks_ai-accounts-fireworks-models-deepseek-v4p1-flash-input_text]": [
       "quota_management.spend_tracking.cost_matrix.logs_cost"
     ],
-    "tests/integration/cost_calculation/test_cost_tracking.py::test_case_bills_expected_cost[fireworks_ai-accounts-fireworks-models-deepseek-v4p1-flash-fallback_cache_read_at_input_rate]": [
+    "tests/integration/cost_calculation/test_cost_tracking.py::test_case_bills_expected_cost[fireworks_ai-accounts-fireworks-models-deepseek-v4p1-flash-fallback_cache_read_at_half_input_rate]": [
       "quota_management.spend_tracking.cost_matrix.logs_cost"
     ],
     "tests/integration/cost_calculation/test_cost_tracking.py::test_case_bills_expected_cost[fireworks_ai-accounts-fireworks-models-deepseek-v4p1-flash-stream]": [
diff --git a/tests/integration/cost_calculation/cost_tracking_cases.json b/tests/integration/cost_calculation/cost_tracking_cases.json
index 428124ac4a9..b0ad270f860 100644
--- a/tests/integration/cost_calculation/cost_tracking_cases.json
+++ b/tests/integration/cost_calculation/cost_tracking_cases.json
@@ -7958,7 +7958,7 @@
       }
     },
     {
-      "name": "fireworks_ai-accounts-fireworks-models-deepseek-v4p1-flash-fallback_cache_read_at_input_rate",
+      "name": "fireworks_ai-accounts-fireworks-models-deepseek-v4p1-flash-fallback_cache_read_at_half_input_rate",
       "covers": "quota_management.spend_tracking.cost_matrix.logs_cost",
       "model": "fireworks_ai/accounts/fireworks/models/deepseek-v4p1-flash",
       "request": {
@@ -8014,9 +8014,10 @@
         }
       },
       "expected": {
-        "spend": 0.0021672,
-        "input_cost": 0.0019392,
+        "spend": 0.0012456,
+        "input_cost": 0.0010176,
         "output_cost": 0.000228,
+        "cache_read_cost": 0.0009216,
         "prompt_tokens": 12928,
         "completion_tokens": 380
       }

From 8907a1d1fccd718d99a5b08f5e9b4750d9536f88 Mon Sep 17 00:00:00 2001
From: kerry <kerry@berri.ai>
Date: Sat, 19 Sep 2026 19:52:43 +0000
Subject: [PATCH 5/6] test: add native responses and messages cost cases

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---
 tests/integration/contracts.json              |   63 +
 .../cost_calculation/cost_tracking_case.py    |    2 +-
 .../cost_calculation/cost_tracking_cases.json | 1079 +++++++++++++++++
 .../cost_calculation/test_cost_tracking.py    |    4 +-
 4 files changed, 1146 insertions(+), 2 deletions(-)

diff --git a/tests/integration/contracts.json b/tests/integration/contracts.json
index 1b0c9e31d4e..04af53a6080 100644
--- a/tests/integration/contracts.json
+++ b/tests/integration/contracts.json
@@ -1320,6 +1320,69 @@
     ],
     "tests/integration/cost_calculation/test_cost_tracking.py::test_case_bills_expected_cost[us.anthropic.claude-opus-5-v1:0-stream_full_usage]": [
       "quota_management.spend_tracking.scripted_wire.logs_cost"
+    ],
+    "tests/integration/cost_calculation/test_cost_tracking.py::test_case_bills_expected_cost[gpt-5.6-responses_cache_read]": [
+      "quota_management.spend_tracking.cost_matrix.logs_cost"
+    ],
+    "tests/integration/cost_calculation/test_cost_tracking.py::test_case_bills_expected_cost[gpt-5.6-responses_reasoning]": [
+      "quota_management.spend_tracking.cost_matrix.logs_cost"
+    ],
+    "tests/integration/cost_calculation/test_cost_tracking.py::test_case_bills_expected_cost[gpt-5.6-responses_stream]": [
+      "quota_management.spend_tracking.scripted_wire.logs_cost"
+    ],
+    "tests/integration/cost_calculation/test_cost_tracking.py::test_case_bills_expected_cost[gpt-5.6-responses_stream_cache_read]": [
+      "quota_management.spend_tracking.scripted_wire.logs_cost"
+    ],
+    "tests/integration/cost_calculation/test_cost_tracking.py::test_case_bills_expected_cost[gpt-5.6-responses_incomplete]": [
+      "quota_management.spend_tracking.cost_matrix.logs_cost"
+    ],
+    "tests/integration/cost_calculation/test_cost_tracking.py::test_case_bills_expected_cost[gpt-5.6-responses_previous_response_id]": [
+      "quota_management.spend_tracking.cost_matrix.logs_cost"
+    ],
+    "tests/integration/cost_calculation/test_cost_tracking.py::test_case_bills_expected_cost[gpt-5.6-responses_web_search_medium]": [
+      "quota_management.spend_tracking.cost_matrix.logs_cost"
+    ],
+    "tests/integration/cost_calculation/test_cost_tracking.py::test_case_bills_expected_cost[gpt-5.3-codex-responses_file_search]": [
+      "quota_management.spend_tracking.cost_matrix.logs_cost"
+    ],
+    "tests/integration/cost_calculation/test_cost_tracking.py::test_case_bills_expected_cost[gpt-5.6-responses_service_tier_flex]": [
+      "quota_management.spend_tracking.cost_matrix.logs_cost"
+    ],
+    "tests/integration/cost_calculation/test_cost_tracking.py::test_case_bills_expected_cost[gpt-5.6-responses_service_tier_priority]": [
+      "quota_management.spend_tracking.cost_matrix.logs_cost"
+    ],
+    "tests/integration/cost_calculation/test_cost_tracking.py::test_case_bills_expected_cost[claude-sonnet-5-messages_input_text]": [
+      "quota_management.spend_tracking.cost_matrix.logs_cost"
+    ],
+    "tests/integration/cost_calculation/test_cost_tracking.py::test_case_bills_expected_cost[claude-sonnet-5-messages_cache_read]": [
+      "quota_management.spend_tracking.cost_matrix.logs_cost"
+    ],
+    "tests/integration/cost_calculation/test_cost_tracking.py::test_case_bills_expected_cost[claude-sonnet-5-messages_cache_write_5m]": [
+      "quota_management.spend_tracking.cost_matrix.logs_cost"
+    ],
+    "tests/integration/cost_calculation/test_cost_tracking.py::test_case_bills_expected_cost[claude-sonnet-5-messages_cache_write_1h]": [
+      "quota_management.spend_tracking.cost_matrix.logs_cost"
+    ],
+    "tests/integration/cost_calculation/test_cost_tracking.py::test_case_bills_expected_cost[claude-sonnet-5-messages_web_search]": [
+      "quota_management.spend_tracking.cost_matrix.logs_cost"
+    ],
+    "tests/integration/cost_calculation/test_cost_tracking.py::test_case_bills_expected_cost[claude-sonnet-5-messages_stream]": [
+      "quota_management.spend_tracking.scripted_wire.logs_cost"
+    ],
+    "tests/integration/cost_calculation/test_cost_tracking.py::test_case_bills_expected_cost[claude-sonnet-5-messages_stream_cache_read]": [
+      "quota_management.spend_tracking.scripted_wire.logs_cost"
+    ],
+    "tests/integration/cost_calculation/test_cost_tracking.py::test_case_bills_expected_cost[claude-sonnet-5-messages_tiered_input_above_200k]": [
+      "quota_management.spend_tracking.cost_matrix.logs_cost"
+    ],
+    "tests/integration/cost_calculation/test_cost_tracking.py::test_case_bills_expected_cost[claude-haiku-4-5-messages_input_text]": [
+      "quota_management.spend_tracking.cost_matrix.logs_cost"
+    ],
+    "tests/integration/cost_calculation/test_cost_tracking.py::test_case_bills_expected_cost[us.anthropic.claude-opus-5-v1:0-messages_input_text]": [
+      "quota_management.spend_tracking.cost_matrix.logs_cost"
+    ],
+    "tests/integration/cost_calculation/test_cost_tracking.py::test_case_bills_expected_cost[anthropic.claude-sonnet-5-v1:0-messages_cache_read]": [
+      "quota_management.spend_tracking.cost_matrix.logs_cost"
     ]
   },
   "browser": {
diff --git a/tests/integration/cost_calculation/cost_tracking_case.py b/tests/integration/cost_calculation/cost_tracking_case.py
index cb408f2c488..830c2999fbf 100644
--- a/tests/integration/cost_calculation/cost_tracking_case.py
+++ b/tests/integration/cost_calculation/cost_tracking_case.py
@@ -172,7 +172,7 @@ class CostTrackingTestCase(BaseModel):
         provider: Final = self.rates.litellm_provider
         prefix: Final = (
             "openai"
-            if provider == "openai" and self.rates.mode == "chat"
+            if provider == "openai" and (self.rates.mode == "chat" or self.endpoint == "/v1/responses")
             else "openai/responses"
             if provider == "openai"
             else _PROVIDER_PREFIXES.get(provider)
diff --git a/tests/integration/cost_calculation/cost_tracking_cases.json b/tests/integration/cost_calculation/cost_tracking_cases.json
index 428124ac4a9..cea9c22577b 100644
--- a/tests/integration/cost_calculation/cost_tracking_cases.json
+++ b/tests/integration/cost_calculation/cost_tracking_cases.json
@@ -25776,6 +25776,1085 @@
         "prompt_tokens": 11056,
         "completion_tokens": 412
       }
+    },
+    {
+      "name": "gpt-5.6-responses_cache_read",
+      "covers": "quota_management.spend_tracking.cost_matrix.logs_cost",
+      "model": "gpt-5.6",
+      "endpoint": "/v1/responses",
+      "request": {
+        "model": "$MODEL",
+        "input": "summarize this text"
+      },
+      "response": {
+        "content_type": "application/json",
+        "body": {
+          "id": "resp_$REQUEST_ID",
+          "object": "response",
+          "created_at": 1700000000,
+          "status": "completed",
+          "model": "gpt-5.6",
+          "output": [
+            {
+              "type": "message",
+              "id": "msg_$REQUEST_ID",
+              "status": "completed",
+              "role": "assistant",
+              "content": [
+                {
+                  "type": "output_text",
+                  "text": "scripted response",
+                  "annotations": []
+                }
+              ]
+            }
+          ],
+          "usage": {
+            "input_tokens": 12928,
+            "output_tokens": 380,
+            "total_tokens": 13308,
+            "input_tokens_details": {
+              "cached_tokens": 12288
+            },
+            "output_tokens_details": {
+              "reasoning_tokens": 0
+            }
+          }
+        }
+      },
+      "expected": {
+        "spend": 0.0085904,
+        "input_cost": 0.0032704,
+        "output_cost": 0.00532,
+        "prompt_tokens": 12928,
+        "completion_tokens": 380,
+        "cache_read_cost": 0.0021504
+      }
+    },
+    {
+      "name": "gpt-5.6-responses_reasoning",
+      "covers": "quota_management.spend_tracking.cost_matrix.logs_cost",
+      "model": "gpt-5.6",
+      "endpoint": "/v1/responses",
+      "request": {
+        "model": "$MODEL",
+        "input": "reason about this text"
+      },
+      "response": {
+        "content_type": "application/json",
+        "body": {
+          "id": "resp_$REQUEST_ID",
+          "object": "response",
+          "created_at": 1700000000,
+          "status": "completed",
+          "model": "gpt-5.6",
+          "output": [
+            {
+              "type": "reasoning",
+              "id": "rs_$REQUEST_ID",
+              "status": "completed",
+              "summary": []
+            },
+            {
+              "type": "message",
+              "id": "msg_$REQUEST_ID",
+              "status": "completed",
+              "role": "assistant",
+              "content": [
+                {
+                  "type": "output_text",
+                  "text": "scripted response",
+                  "annotations": []
+                }
+              ]
+            }
+          ],
+          "usage": {
+            "input_tokens": 1240,
+            "output_tokens": 4040,
+            "total_tokens": 5280,
+            "input_tokens_details": {
+              "cached_tokens": 0
+            },
+            "output_tokens_details": {
+              "reasoning_tokens": 3480
+            }
+          }
+        }
+      },
+      "expected": {
+        "spend": 0.067716,
+        "input_cost": 0.00217,
+        "output_cost": 0.065484,
+        "prompt_tokens": 1240,
+        "completion_tokens": 4040,
+        "reasoning_cost": 0.05742
+      }
+    },
+    {
+      "name": "gpt-5.6-responses_stream",
+      "covers": "quota_management.spend_tracking.scripted_wire.logs_cost",
+      "model": "gpt-5.6",
+      "endpoint": "/v1/responses",
+      "request": {
+        "model": "$MODEL",
+        "input": "stream this text",
+        "stream": true
+      },
+      "response": {
+        "content_type": "text/event-stream",
+        "frames": [
+          "event: response.created\ndata: {\"type\":\"response.created\",\"response\":{\"id\":\"resp_$REQUEST_ID\",\"object\":\"response\",\"created_at\":1700000000,\"status\":\"in_progress\",\"model\":\"gpt-5.6\",\"output\":[{\"type\":\"message\",\"id\":\"msg_$REQUEST_ID\",\"status\":\"completed\",\"role\":\"assistant\",\"content\":[{\"type\":\"output_text\",\"text\":\"scripted response\",\"annotations\":[]}]}],\"usage\":null}}",
+          "event: response.output_item.added\ndata: {\"type\":\"response.output_item.added\",\"output_index\":0,\"item\":{\"type\":\"message\",\"id\":\"msg_$REQUEST_ID\",\"status\":\"in_progress\",\"role\":\"assistant\",\"content\":[]}}",
+          "event: response.content_part.added\ndata: {\"type\":\"response.content_part.added\",\"item_id\":\"msg_$REQUEST_ID\",\"output_index\":0,\"content_index\":0,\"part\":{\"type\":\"content_part\",\"text\":\"\"}}",
+          "event: response.output_text.delta\ndata: {\"type\":\"response.output_text.delta\",\"item_id\":\"msg_$REQUEST_ID\",\"output_index\":0,\"content_index\":0,\"delta\":\"scripted \"}",
+          "event: response.output_text.delta\ndata: {\"type\":\"response.output_text.delta\",\"item_id\":\"msg_$REQUEST_ID\",\"output_index\":0,\"content_index\":0,\"delta\":\"response\"}",
+          "event: response.output_text.done\ndata: {\"type\":\"response.output_text.done\",\"item_id\":\"msg_$REQUEST_ID\",\"output_index\":0,\"content_index\":0,\"text\":\"scripted response\"}",
+          "event: response.content_part.done\ndata: {\"type\":\"response.content_part.done\",\"item_id\":\"msg_$REQUEST_ID\",\"output_index\":0,\"content_index\":0,\"part\":{\"type\":\"output_text\",\"text\":\"scripted response\",\"annotations\":[]}}",
+          "event: response.output_item.done\ndata: {\"type\":\"response.output_item.done\",\"output_index\":0,\"item\":{\"type\":\"message\",\"id\":\"msg_$REQUEST_ID\",\"status\":\"completed\",\"role\":\"assistant\",\"content\":[{\"type\":\"output_text\",\"text\":\"scripted response\",\"annotations\":[]}]}}",
+          "event: response.completed\ndata: {\"type\":\"response.completed\",\"response\":{\"id\":\"resp_$REQUEST_ID\",\"object\":\"response\",\"created_at\":1700000000,\"status\":\"completed\",\"model\":\"gpt-5.6\",\"output\":[{\"type\":\"message\",\"id\":\"msg_$REQUEST_ID\",\"status\":\"completed\",\"role\":\"assistant\",\"content\":[{\"type\":\"output_text\",\"text\":\"scripted response\",\"annotations\":[]}]}],\"usage\":{\"input_tokens\":1840,\"output_tokens\":412,\"total_tokens\":2252,\"input_tokens_details\":{\"cached_tokens\":0},\"output_tokens_details\":{\"reasoning_tokens\":0}}}}"
+        ]
+      },
+      "expected": {
+        "spend": 0.008988,
+        "input_cost": 0.00322,
+        "output_cost": 0.005768,
+        "prompt_tokens": 1840,
+        "completion_tokens": 412
+      }
+    },
+    {
+      "name": "gpt-5.6-responses_stream_cache_read",
+      "covers": "quota_management.spend_tracking.scripted_wire.logs_cost",
+      "model": "gpt-5.6",
+      "endpoint": "/v1/responses",
+      "request": {
+        "model": "$MODEL",
+        "input": "stream cached text",
+        "stream": true
+      },
+      "response": {
+        "content_type": "text/event-stream",
+        "frames": [
+          "event: response.created\ndata: {\"type\":\"response.created\",\"response\":{\"id\":\"resp_$REQUEST_ID\",\"object\":\"response\",\"created_at\":1700000000,\"status\":\"in_progress\",\"model\":\"gpt-5.6\",\"output\":[{\"type\":\"message\",\"id\":\"msg_$REQUEST_ID\",\"status\":\"completed\",\"role\":\"assistant\",\"content\":[{\"type\":\"output_text\",\"text\":\"scripted response\",\"annotations\":[]}]}],\"usage\":null}}",
+          "event: response.output_item.added\ndata: {\"type\":\"response.output_item.added\",\"output_index\":0,\"item\":{\"type\":\"message\",\"id\":\"msg_$REQUEST_ID\",\"status\":\"in_progress\",\"role\":\"assistant\",\"content\":[]}}",
+          "event: response.content_part.added\ndata: {\"type\":\"response.content_part.added\",\"item_id\":\"msg_$REQUEST_ID\",\"output_index\":0,\"content_index\":0,\"part\":{\"type\":\"content_part\",\"text\":\"\"}}",
+          "event: response.output_text.delta\ndata: {\"type\":\"response.output_text.delta\",\"item_id\":\"msg_$REQUEST_ID\",\"output_index\":0,\"content_index\":0,\"delta\":\"scripted \"}",
+          "event: response.output_text.delta\ndata: {\"type\":\"response.output_text.delta\",\"item_id\":\"msg_$REQUEST_ID\",\"output_index\":0,\"content_index\":0,\"delta\":\"response\"}",
+          "event: response.output_text.done\ndata: {\"type\":\"response.output_text.done\",\"item_id\":\"msg_$REQUEST_ID\",\"output_index\":0,\"content_index\":0,\"text\":\"scripted response\"}",
+          "event: response.content_part.done\ndata: {\"type\":\"response.content_part.done\",\"item_id\":\"msg_$REQUEST_ID\",\"output_index\":0,\"content_index\":0,\"part\":{\"type\":\"output_text\",\"text\":\"scripted response\",\"annotations\":[]}}",
+          "event: response.output_item.done\ndata: {\"type\":\"response.output_item.done\",\"output_index\":0,\"item\":{\"type\":\"message\",\"id\":\"msg_$REQUEST_ID\",\"status\":\"completed\",\"role\":\"assistant\",\"content\":[{\"type\":\"output_text\",\"text\":\"scripted response\",\"annotations\":[]}]}}",
+          "event: response.completed\ndata: {\"type\":\"response.completed\",\"response\":{\"id\":\"resp_$REQUEST_ID\",\"object\":\"response\",\"created_at\":1700000000,\"status\":\"completed\",\"model\":\"gpt-5.6\",\"output\":[{\"type\":\"message\",\"id\":\"msg_$REQUEST_ID\",\"status\":\"completed\",\"role\":\"assistant\",\"content\":[{\"type\":\"output_text\",\"text\":\"scripted response\",\"annotations\":[]}]}],\"usage\":{\"input_tokens\":12928,\"output_tokens\":380,\"total_tokens\":13308,\"input_tokens_details\":{\"cached_tokens\":12288},\"output_tokens_details\":{\"reasoning_tokens\":0}}}}"
+        ]
+      },
+      "expected": {
+        "spend": 0.0085904,
+        "input_cost": 0.0032704,
+        "output_cost": 0.00532,
+        "prompt_tokens": 12928,
+        "completion_tokens": 380,
+        "cache_read_cost": 0.0021504
+      }
+    },
+    {
+      "name": "gpt-5.6-responses_incomplete",
+      "covers": "quota_management.spend_tracking.cost_matrix.logs_cost",
+      "model": "gpt-5.6",
+      "endpoint": "/v1/responses",
+      "request": {
+        "model": "$MODEL",
+        "input": "truncate this text",
+        "max_output_tokens": 100
+      },
+      "response": {
+        "content_type": "application/json",
+        "body": {
+          "id": "resp_$REQUEST_ID",
+          "object": "response",
+          "created_at": 1700000000,
+          "status": "incomplete",
+          "model": "gpt-5.6",
+          "output": [
+            {
+              "type": "message",
+              "id": "msg_$REQUEST_ID",
+              "status": "completed",
+              "role": "assistant",
+              "content": [
+                {
+                  "type": "output_text",
+                  "text": "scripted response",
+                  "annotations": []
+                }
+              ]
+            }
+          ],
+          "usage": {
+            "input_tokens": 1840,
+            "output_tokens": 100,
+            "total_tokens": 1940,
+            "input_tokens_details": {
+              "cached_tokens": 0
+            },
+            "output_tokens_details": {
+              "reasoning_tokens": 0
+            }
+          },
+          "incomplete_details": {
+            "reason": "max_output_tokens"
+          }
+        }
+      },
+      "expected": {
+        "spend": 0.00462,
+        "input_cost": 0.00322,
+        "output_cost": 0.0014,
+        "prompt_tokens": 1840,
+        "completion_tokens": 100
+      }
+    },
+    {
+      "name": "gpt-5.6-responses_previous_response_id",
+      "covers": "quota_management.spend_tracking.cost_matrix.logs_cost",
+      "model": "gpt-5.6",
+      "endpoint": "/v1/responses",
+      "request": {
+        "model": "$MODEL",
+        "input": "continue this text",
+        "previous_response_id": "resp_scripted_prior"
+      },
+      "response": {
+        "content_type": "application/json",
+        "body": {
+          "id": "resp_$REQUEST_ID",
+          "object": "response",
+          "created_at": 1700000000,
+          "status": "completed",
+          "model": "gpt-5.6",
+          "output": [
+            {
+              "type": "message",
+              "id": "msg_$REQUEST_ID",
+              "status": "completed",
+              "role": "assistant",
+              "content": [
+                {
+                  "type": "output_text",
+                  "text": "scripted response",
+                  "annotations": []
+                }
+              ]
+            }
+          ],
+          "usage": {
+            "input_tokens": 1840,
+            "output_tokens": 412,
+            "total_tokens": 2252,
+            "input_tokens_details": {
+              "cached_tokens": 0
+            },
+            "output_tokens_details": {
+              "reasoning_tokens": 0
+            }
+          }
+        }
+      },
+      "expected": {
+        "spend": 0.008988,
+        "input_cost": 0.00322,
+        "output_cost": 0.005768,
+        "prompt_tokens": 1840,
+        "completion_tokens": 412
+      }
+    },
+    {
+      "name": "gpt-5.6-responses_web_search_medium",
+      "covers": "quota_management.spend_tracking.cost_matrix.logs_cost",
+      "model": "gpt-5.6",
+      "endpoint": "/v1/responses",
+      "request": {
+        "model": "$MODEL",
+        "input": "search this text",
+        "tools": [
+          {
+            "type": "web_search_preview",
+            "search_context_size": "medium"
+          }
+        ]
+      },
+      "response": {
+        "content_type": "application/json",
+        "body": {
+          "id": "resp_$REQUEST_ID",
+          "object": "response",
+          "created_at": 1700000000,
+          "status": "completed",
+          "model": "gpt-5.6",
+          "output": [
+            {
+              "type": "web_search_call",
+              "id": "ws_$REQUEST_ID",
+              "status": "completed",
+              "action": {
+                "type": "search",
+                "query": "scripted query"
+              }
+            },
+            {
+              "type": "message",
+              "id": "msg_$REQUEST_ID",
+              "status": "completed",
+              "role": "assistant",
+              "content": [
+                {
+                  "type": "output_text",
+                  "text": "scripted response",
+                  "annotations": []
+                }
+              ]
+            }
+          ],
+          "usage": {
+            "input_tokens": 1840,
+            "output_tokens": 412,
+            "total_tokens": 2252,
+            "input_tokens_details": {
+              "cached_tokens": 0
+            },
+            "output_tokens_details": {
+              "reasoning_tokens": 0
+            }
+          }
+        }
+      },
+      "expected": {
+        "spend": 0.021488,
+        "input_cost": 0.00322,
+        "output_cost": 0.005768,
+        "prompt_tokens": 1840,
+        "completion_tokens": 412,
+        "tool_usage_cost": 0.0125
+      }
+    },
+    {
+      "name": "gpt-5.3-codex-responses_file_search",
+      "covers": "quota_management.spend_tracking.cost_matrix.logs_cost",
+      "model": "gpt-5.3-codex",
+      "endpoint": "/v1/responses",
+      "request": {
+        "model": "$MODEL",
+        "input": "search files",
+        "tools": [
+          {
+            "type": "file_search",
+            "vector_store_ids": [
+              "vs_scripted"
+            ]
+          }
+        ]
+      },
+      "response": {
+        "content_type": "application/json",
+        "body": {
+          "id": "resp_$REQUEST_ID",
+          "object": "response",
+          "created_at": 1700000000,
+          "status": "completed",
+          "model": "gpt-5.3-codex",
+          "output": [
+            {
+              "type": "file_search_call",
+              "id": "fs_$REQUEST_ID",
+              "status": "completed",
+              "queries": [
+                "scripted query"
+              ],
+              "results": []
+            },
+            {
+              "type": "message",
+              "id": "msg_$REQUEST_ID",
+              "status": "completed",
+              "role": "assistant",
+              "content": [
+                {
+                  "type": "output_text",
+                  "text": "scripted response",
+                  "annotations": []
+                }
+              ]
+            }
+          ],
+          "usage": {
+            "input_tokens": 1840,
+            "output_tokens": 412,
+            "total_tokens": 2252,
+            "input_tokens_details": {
+              "cached_tokens": 0
+            },
+            "output_tokens_details": {
+              "reasoning_tokens": 0
+            }
+          }
+        }
+      },
+      "expected": {
+        "spend": 0.010204,
+        "input_cost": 0.00276,
+        "output_cost": 0.004944,
+        "prompt_tokens": 1840,
+        "completion_tokens": 412,
+        "tool_usage_cost": 0.0025
+      }
+    },
+    {
+      "name": "gpt-5.6-responses_service_tier_flex",
+      "covers": "quota_management.spend_tracking.cost_matrix.logs_cost",
+      "model": "gpt-5.6",
+      "endpoint": "/v1/responses",
+      "request": {
+        "model": "$MODEL",
+        "input": "flex text",
+        "service_tier": "flex"
+      },
+      "response": {
+        "content_type": "application/json",
+        "body": {
+          "id": "resp_$REQUEST_ID",
+          "object": "response",
+          "created_at": 1700000000,
+          "status": "completed",
+          "model": "gpt-5.6",
+          "output": [
+            {
+              "type": "message",
+              "id": "msg_$REQUEST_ID",
+              "status": "completed",
+              "role": "assistant",
+              "content": [
+                {
+                  "type": "output_text",
+                  "text": "scripted response",
+                  "annotations": []
+                }
+              ]
+            }
+          ],
+          "usage": {
+            "input_tokens": 1840,
+            "output_tokens": 412,
+            "total_tokens": 2252,
+            "input_tokens_details": {
+              "cached_tokens": 0
+            },
+            "output_tokens_details": {
+              "reasoning_tokens": 0
+            },
+            "service_tier": "flex"
+          },
+          "service_tier": "flex"
+        }
+      },
+      "expected": {
+        "spend": 0.004494,
+        "input_cost": 0.00161,
+        "output_cost": 0.002884,
+        "prompt_tokens": 1840,
+        "completion_tokens": 412
+      }
+    },
+    {
+      "name": "gpt-5.6-responses_service_tier_priority",
+      "covers": "quota_management.spend_tracking.cost_matrix.logs_cost",
+      "model": "gpt-5.6",
+      "endpoint": "/v1/responses",
+      "request": {
+        "model": "$MODEL",
+        "input": "priority text",
+        "service_tier": "priority"
+      },
+      "response": {
+        "content_type": "application/json",
+        "body": {
+          "id": "resp_$REQUEST_ID",
+          "object": "response",
+          "created_at": 1700000000,
+          "status": "completed",
+          "model": "gpt-5.6",
+          "output": [
+            {
+              "type": "message",
+              "id": "msg_$REQUEST_ID",
+              "status": "completed",
+              "role": "assistant",
+              "content": [
+                {
+                  "type": "output_text",
+                  "text": "scripted response",
+                  "annotations": []
+                }
+              ]
+            }
+          ],
+          "usage": {
+            "input_tokens": 1840,
+            "output_tokens": 412,
+            "total_tokens": 2252,
+            "input_tokens_details": {
+              "cached_tokens": 0
+            },
+            "output_tokens_details": {
+              "reasoning_tokens": 0
+            },
+            "service_tier": "priority"
+          },
+          "service_tier": "priority"
+        }
+      },
+      "expected": {
+        "spend": 0.017976,
+        "input_cost": 0.00644,
+        "output_cost": 0.011536,
+        "prompt_tokens": 1840,
+        "completion_tokens": 412
+      }
+    },
+    {
+      "name": "claude-sonnet-5-messages_input_text",
+      "covers": "quota_management.spend_tracking.cost_matrix.logs_cost",
+      "model": "claude-sonnet-5",
+      "endpoint": "/v1/messages",
+      "request": {
+        "model": "$MODEL",
+        "max_tokens": 412,
+        "messages": [
+          {
+            "role": "user",
+            "content": "summarize the attached material in one line"
+          }
+        ]
+      },
+      "response": {
+        "content_type": "application/json",
+        "body": {
+          "id": "msg_$REQUEST_ID",
+          "type": "message",
+          "role": "assistant",
+          "model": "claude-sonnet-5",
+          "content": [
+            {
+              "type": "text",
+              "text": "scripted response"
+            }
+          ],
+          "stop_reason": "end_turn",
+          "usage": {
+            "input_tokens": 1840,
+            "output_tokens": 412
+          }
+        }
+      },
+      "expected": {
+        "spend": 0.0117,
+        "input_cost": 0.00552,
+        "output_cost": 0.00618,
+        "prompt_tokens": 1840,
+        "completion_tokens": 412
+      }
+    },
+    {
+      "name": "claude-sonnet-5-messages_cache_read",
+      "covers": "quota_management.spend_tracking.cost_matrix.logs_cost",
+      "model": "claude-sonnet-5",
+      "endpoint": "/v1/messages",
+      "request": {
+        "model": "$MODEL",
+        "max_tokens": 412,
+        "messages": [
+          {
+            "role": "user",
+            "content": [
+              {
+                "type": "text",
+                "text": "cached text",
+                "cache_control": {
+                  "type": "ephemeral"
+                }
+              }
+            ]
+          }
+        ]
+      },
+      "response": {
+        "content_type": "application/json",
+        "body": {
+          "id": "msg_$REQUEST_ID",
+          "type": "message",
+          "role": "assistant",
+          "model": "claude-sonnet-5",
+          "content": [
+            {
+              "type": "text",
+              "text": "scripted response"
+            }
+          ],
+          "stop_reason": "end_turn",
+          "usage": {
+            "input_tokens": 640,
+            "output_tokens": 380,
+            "cache_read_input_tokens": 12288
+          }
+        }
+      },
+      "expected": {
+        "spend": 0.0113064,
+        "input_cost": 0.0056064,
+        "output_cost": 0.0057,
+        "prompt_tokens": 12928,
+        "completion_tokens": 380,
+        "cache_read_cost": 0.0036864
+      }
+    },
+    {
+      "name": "claude-sonnet-5-messages_cache_write_5m",
+      "covers": "quota_management.spend_tracking.cost_matrix.logs_cost",
+      "model": "claude-sonnet-5",
+      "endpoint": "/v1/messages",
+      "request": {
+        "model": "$MODEL",
+        "max_tokens": 350,
+        "messages": [
+          {
+            "role": "user",
+            "content": [
+              {
+                "type": "text",
+                "text": "cache this text",
+                "cache_control": {
+                  "type": "ephemeral"
+                }
+              }
+            ]
+          }
+        ]
+      },
+      "response": {
+        "content_type": "application/json",
+        "body": {
+          "id": "msg_$REQUEST_ID",
+          "type": "message",
+          "role": "assistant",
+          "model": "claude-sonnet-5",
+          "content": [
+            {
+              "type": "text",
+              "text": "scripted response"
+            }
+          ],
+          "stop_reason": "end_turn",
+          "usage": {
+            "input_tokens": 512,
+            "output_tokens": 350,
+            "cache_creation_input_tokens": 9216,
+            "cache_creation": {
+              "ephemeral_5m_input_tokens": 9216,
+              "ephemeral_1h_input_tokens": 0
+            }
+          }
+        }
+      },
+      "expected": {
+        "spend": 0.041346,
+        "input_cost": 0.036096,
+        "output_cost": 0.00525,
+        "prompt_tokens": 9728,
+        "completion_tokens": 350,
+        "cache_creation_cost": 0.03456
+      }
+    },
+    {
+      "name": "claude-sonnet-5-messages_cache_write_1h",
+      "covers": "quota_management.spend_tracking.cost_matrix.logs_cost",
+      "model": "claude-sonnet-5",
+      "endpoint": "/v1/messages",
+      "request": {
+        "model": "$MODEL",
+        "max_tokens": 350,
+        "messages": [
+          {
+            "role": "user",
+            "content": [
+              {
+                "type": "text",
+                "text": "cache this text for an hour",
+                "cache_control": {
+                  "type": "ephemeral",
+                  "ttl": "1h"
+                }
+              }
+            ]
+          }
+        ]
+      },
+      "response": {
+        "content_type": "application/json",
+        "body": {
+          "id": "msg_$REQUEST_ID",
+          "type": "message",
+          "role": "assistant",
+          "model": "claude-sonnet-5",
+          "content": [
+            {
+              "type": "text",
+              "text": "scripted response"
+            }
+          ],
+          "stop_reason": "end_turn",
+          "usage": {
+            "input_tokens": 512,
+            "output_tokens": 350,
+            "cache_creation_input_tokens": 9216,
+            "cache_creation": {
+              "ephemeral_5m_input_tokens": 2048,
+              "ephemeral_1h_input_tokens": 7168
+            }
+          }
+        }
+      },
+      "expected": {
+        "spend": 0.057474,
+        "input_cost": 0.052224,
+        "output_cost": 0.00525,
+        "prompt_tokens": 9728,
+        "completion_tokens": 350,
+        "cache_creation_cost": 0.050688
+      }
+    },
+    {
+      "name": "claude-sonnet-5-messages_web_search",
+      "covers": "quota_management.spend_tracking.cost_matrix.logs_cost",
+      "model": "claude-sonnet-5",
+      "endpoint": "/v1/messages",
+      "request": {
+        "model": "$MODEL",
+        "max_tokens": 412,
+        "messages": [
+          {
+            "role": "user",
+            "content": "summarize the attached material in one line"
+          }
+        ],
+        "tools": [
+          {
+            "type": "web_search_20250305",
+            "name": "web_search"
+          }
+        ]
+      },
+      "response": {
+        "content_type": "application/json",
+        "body": {
+          "id": "msg_$REQUEST_ID",
+          "type": "message",
+          "role": "assistant",
+          "model": "claude-sonnet-5",
+          "content": [
+            {
+              "type": "server_tool_use",
+              "id": "srv_$REQUEST_ID",
+              "name": "web_search",
+              "input": {
+                "query": "scripted query"
+              }
+            },
+            {
+              "type": "web_search_tool_result",
+              "tool_use_id": "srv_$REQUEST_ID",
+              "content": [
+                {
+                  "type": "web_search_result",
+                  "title": "scripted result",
+                  "url": "https://scripted.example"
+                }
+              ]
+            },
+            {
+              "type": "text",
+              "text": "scripted response"
+            }
+          ],
+          "stop_reason": "end_turn",
+          "usage": {
+            "input_tokens": 1840,
+            "output_tokens": 412,
+            "server_tool_use": {
+              "web_search_requests": 2
+            }
+          }
+        }
+      },
+      "expected": {
+        "spend": 0.0317,
+        "input_cost": 0.00552,
+        "output_cost": 0.00618,
+        "prompt_tokens": 1840,
+        "completion_tokens": 412,
+        "tool_usage_cost": 0.02
+      }
+    },
+    {
+      "name": "claude-sonnet-5-messages_stream",
+      "covers": "quota_management.spend_tracking.scripted_wire.logs_cost",
+      "model": "claude-sonnet-5",
+      "endpoint": "/v1/messages",
+      "request": {
+        "model": "$MODEL",
+        "max_tokens": 412,
+        "messages": [
+          {
+            "role": "user",
+            "content": "summarize the attached material in one line"
+          }
+        ],
+        "stream": true
+      },
+      "response": {
+        "content_type": "text/event-stream",
+        "frames": [
+          "event: message_start\ndata: {\"type\":\"message_start\",\"message\":{\"id\":\"msg_$REQUEST_ID\",\"type\":\"message\",\"role\":\"assistant\",\"model\":\"claude-sonnet-5\",\"content\":[],\"stop_reason\":null,\"usage\":{\"input_tokens\":1840}}}",
+          "event: content_block_start\ndata: {\"type\":\"content_block_start\",\"index\":0,\"content_block\":{\"type\":\"text\",\"text\":\"\"}}",
+          "event: content_block_delta\ndata: {\"type\":\"content_block_delta\",\"index\":0,\"delta\":{\"type\":\"text_delta\",\"text\":\"scripted \"}}",
+          "event: content_block_delta\ndata: {\"type\":\"content_block_delta\",\"index\":0,\"delta\":{\"type\":\"text_delta\",\"text\":\"response\"}}",
+          "event: content_block_stop\ndata: {\"type\":\"content_block_stop\",\"index\":0}",
+          "event: message_delta\ndata: {\"type\":\"message_delta\",\"delta\":{\"stop_reason\":\"end_turn\"},\"usage\":{\"output_tokens\":412}}",
+          "event: message_stop\ndata: {\"type\":\"message_stop\"}"
+        ]
+      },
+      "expected": {
+        "spend": 0.0117,
+        "input_cost": 0.00552,
+        "output_cost": 0.00618,
+        "prompt_tokens": 1840,
+        "completion_tokens": 412
+      }
+    },
+    {
+      "name": "claude-sonnet-5-messages_stream_cache_read",
+      "covers": "quota_management.spend_tracking.scripted_wire.logs_cost",
+      "model": "claude-sonnet-5",
+      "endpoint": "/v1/messages",
+      "request": {
+        "model": "$MODEL",
+        "max_tokens": 380,
+        "messages": [
+          {
+            "role": "user",
+            "content": "summarize the attached material in one line"
+          }
+        ],
+        "stream": true
+      },
+      "response": {
+        "content_type": "text/event-stream",
+        "frames": [
+          "event: message_start\ndata: {\"type\":\"message_start\",\"message\":{\"id\":\"msg_$REQUEST_ID\",\"type\":\"message\",\"role\":\"assistant\",\"model\":\"claude-sonnet-5\",\"content\":[],\"stop_reason\":null,\"usage\":{\"input_tokens\":640,\"cache_read_input_tokens\":12288}}}",
+          "event: content_block_start\ndata: {\"type\":\"content_block_start\",\"index\":0,\"content_block\":{\"type\":\"text\",\"text\":\"\"}}",
+          "event: content_block_delta\ndata: {\"type\":\"content_block_delta\",\"index\":0,\"delta\":{\"type\":\"text_delta\",\"text\":\"scripted \"}}",
+          "event: content_block_delta\ndata: {\"type\":\"content_block_delta\",\"index\":0,\"delta\":{\"type\":\"text_delta\",\"text\":\"response\"}}",
+          "event: content_block_stop\ndata: {\"type\":\"content_block_stop\",\"index\":0}",
+          "event: message_delta\ndata: {\"type\":\"message_delta\",\"delta\":{\"stop_reason\":\"end_turn\"},\"usage\":{\"output_tokens\":380}}",
+          "event: message_stop\ndata: {\"type\":\"message_stop\"}"
+        ]
+      },
+      "expected": {
+        "spend": 0.0113064,
+        "input_cost": 0.0056064,
+        "output_cost": 0.0057,
+        "prompt_tokens": 12928,
+        "completion_tokens": 380,
+        "cache_read_cost": 0.0036864
+      }
+    },
+    {
+      "name": "claude-sonnet-5-messages_tiered_input_above_200k",
+      "covers": "quota_management.spend_tracking.cost_matrix.logs_cost",
+      "model": "claude-sonnet-5",
+      "endpoint": "/v1/messages",
+      "request": {
+        "model": "$MODEL",
+        "max_tokens": 620,
+        "messages": [
+          {
+            "role": "user",
+            "content": "summarize the attached material in one line"
+          }
+        ]
+      },
+      "response": {
+        "content_type": "application/json",
+        "body": {
+          "id": "msg_$REQUEST_ID",
+          "type": "message",
+          "role": "assistant",
+          "model": "claude-sonnet-5",
+          "content": [
+            {
+              "type": "text",
+              "text": "scripted response"
+            }
+          ],
+          "stop_reason": "end_turn",
+          "usage": {
+            "input_tokens": 210000,
+            "output_tokens": 620
+          }
+        }
+      },
+      "expected": {
+        "spend": 1.2693,
+        "input_cost": 1.26,
+        "output_cost": 0.0093,
+        "prompt_tokens": 210000,
+        "completion_tokens": 620
+      }
+    },
+    {
+      "name": "claude-haiku-4-5-messages_input_text",
+      "covers": "quota_management.spend_tracking.cost_matrix.logs_cost",
+      "model": "claude-haiku-4-5",
+      "endpoint": "/v1/messages",
+      "request": {
+        "model": "$MODEL",
+        "max_tokens": 412,
+        "messages": [
+          {
+            "role": "user",
+            "content": "summarize the attached material in one line"
+          }
+        ]
+      },
+      "response": {
+        "content_type": "application/json",
+        "body": {
+          "id": "msg_$REQUEST_ID",
+          "type": "message",
+          "role": "assistant",
+          "model": "claude-haiku-4-5",
+          "content": [
+            {
+              "type": "text",
+              "text": "scripted response"
+            }
+          ],
+          "stop_reason": "end_turn",
+          "usage": {
+            "input_tokens": 1840,
+            "output_tokens": 412
+          }
+        }
+      },
+      "expected": {
+        "spend": 0.0039,
+        "input_cost": 0.00184,
+        "output_cost": 0.00206,
+        "prompt_tokens": 1840,
+        "completion_tokens": 412
+      }
+    },
+    {
+      "name": "us.anthropic.claude-opus-5-v1:0-messages_input_text",
+      "covers": "quota_management.spend_tracking.cost_matrix.logs_cost",
+      "model": "us.anthropic.claude-opus-5-v1:0",
+      "endpoint": "/v1/messages",
+      "request": {
+        "model": "$MODEL",
+        "max_tokens": 412,
+        "messages": [
+          {
+            "role": "user",
+            "content": "summarize the attached material in one line"
+          }
+        ]
+      },
+      "response": {
+        "content_type": "application/json",
+        "body": {
+          "output": {
+            "message": {
+              "role": "assistant",
+              "content": [
+                {
+                  "text": "scripted response"
+                }
+              ]
+            }
+          },
+          "stopReason": "end_turn",
+          "usage": {
+            "inputTokens": 1840,
+            "outputTokens": 412
+          },
+          "metrics": {
+            "latencyMs": 42
+          }
+        }
+      },
+      "expected": {
+        "spend": 0.02145,
+        "input_cost": 0.01012,
+        "output_cost": 0.01133,
+        "prompt_tokens": 1840,
+        "completion_tokens": 412
+      }
+    },
+    {
+      "name": "anthropic.claude-sonnet-5-v1:0-messages_cache_read",
+      "covers": "quota_management.spend_tracking.cost_matrix.logs_cost",
+      "model": "anthropic.claude-sonnet-5-v1:0",
+      "endpoint": "/v1/messages",
+      "request": {
+        "model": "$MODEL",
+        "max_tokens": 412,
+        "messages": [
+          {
+            "role": "user",
+            "content": "summarize the attached material in one line"
+          }
+        ]
+      },
+      "response": {
+        "content_type": "application/json",
+        "body": {
+          "output": {
+            "message": {
+              "role": "assistant",
+              "content": [
+                {
+                  "text": "scripted response"
+                }
+              ]
+            }
+          },
+          "stopReason": "end_turn",
+          "usage": {
+            "inputTokens": 640,
+            "outputTokens": 380,
+            "totalTokens": 13308,
+            "cacheReadInputTokens": 12288
+          },
+          "metrics": {
+            "latencyMs": 42
+          }
+        }
+      },
+      "expected": {
+        "spend": 0.01243704,
+        "input_cost": 0.00616704,
+        "output_cost": 0.00627,
+        "prompt_tokens": 12928,
+        "completion_tokens": 380,
+        "cache_read_cost": 0.00405504
+      }
     }
   ]
 }
diff --git a/tests/integration/cost_calculation/test_cost_tracking.py b/tests/integration/cost_calculation/test_cost_tracking.py
index 5b876974c63..fd18c400b4e 100644
--- a/tests/integration/cost_calculation/test_cost_tracking.py
+++ b/tests/integration/cost_calculation/test_cost_tracking.py
@@ -42,7 +42,9 @@ def _assert_stream_has_no_error(response_text: str) -> None:
         if payload == "[DONE]":
             continue
         parsed = JSON_OBJECT.validate_json(payload)
-        assert "error" not in parsed, f"stream carried an error event: {parsed}"
+        assert (
+            "error" not in parsed and parsed.get("type") not in {"error", "response.failed"}
+        ), f"stream carried an error event: {parsed}"
 
 
 @pytest.mark.parametrize("case", _CASES)

From ca8d0e500c56018ce1a9a4dae54f741ffc6edfe2 Mon Sep 17 00:00:00 2001
From: kerry <kerry@berri.ai>
Date: Sat, 19 Sep 2026 20:11:24 +0000
Subject: [PATCH 6/6] test(integration): correct responses reasoning and
 messages tiered expectations

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---
 .../cost_calculation/cost_tracking_cases.json | 21 ++++++++++++-------
 1 file changed, 14 insertions(+), 7 deletions(-)

diff --git a/tests/integration/cost_calculation/cost_tracking_cases.json b/tests/integration/cost_calculation/cost_tracking_cases.json
index 8f36e679e9f..ac4c6c125fb 100644
--- a/tests/integration/cost_calculation/cost_tracking_cases.json
+++ b/tests/integration/cost_calculation/cost_tracking_cases.json
@@ -25884,12 +25884,12 @@
         }
       },
       "expected": {
-        "spend": 0.067716,
+        "spend": 0.06569,
         "input_cost": 0.00217,
-        "output_cost": 0.065484,
+        "output_cost": 0.06352,
         "prompt_tokens": 1240,
         "completion_tokens": 4040,
-        "reasoning_cost": 0.05742
+        "reasoning_cost": 0.05568
       }
     },
     {
@@ -26349,6 +26349,7 @@
             }
           ],
           "stop_reason": "end_turn",
+          "stop_sequence": null,
           "usage": {
             "input_tokens": 1840,
             "output_tokens": 412
@@ -26400,6 +26401,7 @@
             }
           ],
           "stop_reason": "end_turn",
+          "stop_sequence": null,
           "usage": {
             "input_tokens": 640,
             "output_tokens": 380,
@@ -26453,6 +26455,7 @@
             }
           ],
           "stop_reason": "end_turn",
+          "stop_sequence": null,
           "usage": {
             "input_tokens": 512,
             "output_tokens": 350,
@@ -26511,6 +26514,7 @@
             }
           ],
           "stop_reason": "end_turn",
+          "stop_sequence": null,
           "usage": {
             "input_tokens": 512,
             "output_tokens": 350,
@@ -26585,6 +26589,7 @@
             }
           ],
           "stop_reason": "end_turn",
+          "stop_sequence": null,
           "usage": {
             "input_tokens": 1840,
             "output_tokens": 412,
@@ -26622,7 +26627,7 @@
       "response": {
         "content_type": "text/event-stream",
         "frames": [
-          "event: message_start\ndata: {\"type\":\"message_start\",\"message\":{\"id\":\"msg_$REQUEST_ID\",\"type\":\"message\",\"role\":\"assistant\",\"model\":\"claude-sonnet-5\",\"content\":[],\"stop_reason\":null,\"usage\":{\"input_tokens\":1840}}}",
+          "event: message_start\ndata: {\"type\":\"message_start\",\"message\":{\"id\":\"msg_$REQUEST_ID\",\"type\":\"message\",\"role\":\"assistant\",\"model\":\"claude-sonnet-5\",\"content\":[],\"stop_reason\":null,\"stop_sequence\":null,\"usage\":{\"input_tokens\":1840}}}",
           "event: content_block_start\ndata: {\"type\":\"content_block_start\",\"index\":0,\"content_block\":{\"type\":\"text\",\"text\":\"\"}}",
           "event: content_block_delta\ndata: {\"type\":\"content_block_delta\",\"index\":0,\"delta\":{\"type\":\"text_delta\",\"text\":\"scripted \"}}",
           "event: content_block_delta\ndata: {\"type\":\"content_block_delta\",\"index\":0,\"delta\":{\"type\":\"text_delta\",\"text\":\"response\"}}",
@@ -26658,7 +26663,7 @@
       "response": {
         "content_type": "text/event-stream",
         "frames": [
-          "event: message_start\ndata: {\"type\":\"message_start\",\"message\":{\"id\":\"msg_$REQUEST_ID\",\"type\":\"message\",\"role\":\"assistant\",\"model\":\"claude-sonnet-5\",\"content\":[],\"stop_reason\":null,\"usage\":{\"input_tokens\":640,\"cache_read_input_tokens\":12288}}}",
+          "event: message_start\ndata: {\"type\":\"message_start\",\"message\":{\"id\":\"msg_$REQUEST_ID\",\"type\":\"message\",\"role\":\"assistant\",\"model\":\"claude-sonnet-5\",\"content\":[],\"stop_reason\":null,\"stop_sequence\":null,\"usage\":{\"input_tokens\":640,\"cache_read_input_tokens\":12288}}}",
           "event: content_block_start\ndata: {\"type\":\"content_block_start\",\"index\":0,\"content_block\":{\"type\":\"text\",\"text\":\"\"}}",
           "event: content_block_delta\ndata: {\"type\":\"content_block_delta\",\"index\":0,\"delta\":{\"type\":\"text_delta\",\"text\":\"scripted \"}}",
           "event: content_block_delta\ndata: {\"type\":\"content_block_delta\",\"index\":0,\"delta\":{\"type\":\"text_delta\",\"text\":\"response\"}}",
@@ -26705,6 +26710,7 @@
             }
           ],
           "stop_reason": "end_turn",
+          "stop_sequence": null,
           "usage": {
             "input_tokens": 210000,
             "output_tokens": 620
@@ -26712,9 +26718,9 @@
         }
       },
       "expected": {
-        "spend": 1.2693,
+        "spend": 1.27395,
         "input_cost": 1.26,
-        "output_cost": 0.0093,
+        "output_cost": 0.01395,
         "prompt_tokens": 210000,
         "completion_tokens": 620
       }
@@ -26748,6 +26754,7 @@
             }
           ],
           "stop_reason": "end_turn",
+          "stop_sequence": null,
           "usage": {
             "input_tokens": 1840,
             "output_tokens": 412