From 65e2432a370bd4eb162a5a4c2325200e1dc90399 Mon Sep 17 00:00:00 2001
From: Shivam Rawat <shivamrawat@Shivams-MacBook-Pro.local>
Date: Sat, 11 Jul 2026 13:47:56 -0700
Subject: [PATCH] fix(proxy): bill tier-only deployments instead of $0

Route cost calculation to the deployment's router_model_id entry when it carries tiered_pricing but no flat per-token rate, so models like dashscope/qwen3.7-plus are billed via their tier table rather than the pricing-stripped shared alias.

Co-authored-by: Cursor <cursoragent@cursor.com>
---
 litellm/cost_calculator.py                 |   6 +-
 tests/test_litellm/test_cost_calculator.py | 104 +++++++++++++++++++++
 2 files changed, 109 insertions(+), 1 deletion(-)

diff --git a/litellm/cost_calculator.py b/litellm/cost_calculator.py
index 74dc0e19da3..a40a8e1389c 100644
--- a/litellm/cost_calculator.py
+++ b/litellm/cost_calculator.py
@@ -760,7 +760,11 @@ def _select_model_name_for_cost_calc(
     if custom_pricing is True:
         if router_model_id is not None and router_model_id in litellm.model_cost:
             entry = litellm.model_cost[router_model_id]
-            if entry.get("input_cost_per_token") is not None or entry.get("input_cost_per_second") is not None:
+            if (
+                entry.get("input_cost_per_token") is not None
+                or entry.get("input_cost_per_second") is not None
+                or entry.get("tiered_pricing") is not None
+            ):
                 return_model = router_model_id
             else:
                 return_model = model
diff --git a/tests/test_litellm/test_cost_calculator.py b/tests/test_litellm/test_cost_calculator.py
index 751177014ce..9636db4f4cd 100644
--- a/tests/test_litellm/test_cost_calculator.py
+++ b/tests/test_litellm/test_cost_calculator.py
@@ -1000,6 +1000,110 @@ def test_per_request_custom_pricing_with_router():
     assert "gpt-3.5-turbo" in selected
 
 
+def test_tiered_pricing_only_deployment_selects_router_model_id():
+    """A deployment priced solely via ``tiered_pricing`` (no flat
+    input/output cost) must resolve cost against its ``router_model_id``
+    entry, which holds the tiered table, instead of the shared backend alias
+    that has custom pricing fields stripped. Regression for tier-only models
+    (e.g. dashscope/qwen3.7-plus) being billed as free.
+    """
+    from litellm import Router
+    from litellm.cost_calculator import _select_model_name_for_cost_calc
+
+    router = Router(
+        model_list=[
+            {
+                "model_name": "qwen-3.7-plus",
+                "litellm_params": {
+                    "model": "dashscope/qwen3.7-plus",
+                    "api_key": "sk-fake",
+                },
+                "model_info": {
+                    "tiered_pricing": [
+                        {
+                            "input_cost_per_token": 4e-07,
+                            "output_cost_per_token": 1.6e-06,
+                            "range": [0, 256000],
+                        },
+                    ],
+                },
+            },
+        ]
+    )
+    router_model_id = router.model_list[0]["model_info"]["id"]
+
+    entry = litellm.model_cost[router_model_id]
+    assert entry.get("input_cost_per_token") is None
+    assert entry.get("tiered_pricing") is not None
+    # The stripped shared alias must not carry tiered pricing.
+    assert litellm.model_cost["dashscope/qwen3.7-plus"].get("tiered_pricing") is None
+
+    selected = _select_model_name_for_cost_calc(
+        model="dashscope/qwen3.7-plus",
+        completion_response=None,
+        custom_pricing=True,
+        custom_llm_provider="dashscope",
+        router_model_id=router_model_id,
+    )
+    assert selected is not None
+    assert router_model_id in selected
+
+
+def test_tiered_pricing_only_deployment_completion_cost_is_nonzero():
+    """End-to-end: a tier-only deployment must produce the tiered cost, not
+    $0. Mirrors the reported dashscope/qwen3.7-plus trace (12 prompt + 377
+    completion tokens).
+    """
+    from litellm import Router
+    from litellm.types.utils import Choices, Message
+
+    router = Router(
+        model_list=[
+            {
+                "model_name": "qwen-3.7-plus",
+                "litellm_params": {
+                    "model": "dashscope/qwen3.7-plus",
+                    "api_key": "sk-fake",
+                },
+                "model_info": {
+                    "tiered_pricing": [
+                        {
+                            "input_cost_per_token": 4e-07,
+                            "output_cost_per_token": 1.6e-06,
+                            "range": [0, 256000],
+                        },
+                        {
+                            "input_cost_per_token": 1.2e-06,
+                            "output_cost_per_token": 4.8e-06,
+                            "range": [256000, 1000000],
+                        },
+                    ],
+                },
+            },
+        ]
+    )
+    router_model_id = router.model_list[0]["model_info"]["id"]
+
+    response = ModelResponse(
+        model="dashscope/qwen3.7-plus",
+        choices=[Choices(index=0, message=Message(role="assistant", content="hi"))],
+        usage=Usage(prompt_tokens=12, completion_tokens=377, total_tokens=389),
+    )
+    response._hidden_params = {"custom_llm_provider": "dashscope", "model_id": router_model_id}
+
+    cost = completion_cost(
+        completion_response=response,
+        model="dashscope/qwen3.7-plus",
+        custom_llm_provider="dashscope",
+        custom_pricing=True,
+        router_model_id=router_model_id,
+    )
+
+    expected = 12 * 4e-07 + 377 * 1.6e-06
+    assert cost == pytest.approx(expected)
+    assert cost > 0
+
+
 def test_azure_realtime_cost_calculator():
     os.environ["LITELLM_LOCAL_MODEL_COST_MAP"] = "True"
     litellm.model_cost = litellm.get_model_cost_map(url="")