From aad96a42d3202d879c9fbbb94b406c8e38fab1a0 Mon Sep 17 00:00:00 2001
From: mateo-berri <277851410+mateo-berri@users.noreply.github.com>
Date: Wed, 28 Oct 2026 07:29:16 -0700
Subject: [PATCH] fix(cost): apply the priced fallback to the provider response
 model streaming records

---
 litellm/cost_calculator.py         | 23 ++++++++++++++---------
 tests/unit/test_cost_calculator.py | 20 ++++++++++++++++++++
 2 files changed, 34 insertions(+), 9 deletions(-)

diff --git a/litellm/cost_calculator.py b/litellm/cost_calculator.py
index 43f86981707..9ac414a3d29 100644
--- a/litellm/cost_calculator.py
+++ b/litellm/cost_calculator.py
@@ -996,19 +996,18 @@ def select_model_name_for_cost_calc(
         else:
             return_model = model
 
-    elif base_model is not None or bedrock_pricing_model is not None or provider_response_model is not None:
-        return_model = base_model if base_model is not None else bedrock_pricing_model or provider_response_model
+    elif base_model is not None or bedrock_pricing_model is not None:
+        return_model = base_model if base_model is not None else bedrock_pricing_model
+
+    elif provider_response_model is not None:
+        return _served_or_priced_requested_model(provider_response_model, model, custom_llm_provider, priced_region)
 
     elif completion_response_model is None and hidden_params is not None:
         if hidden_params.get("model", None) is not None and len(hidden_params["model"]) > 0:
             return_model = hidden_params.get("model", model)
 
     if return_model is None and completion_response_model is not None:
-        return _served_or_priced_requested_model(
-            served=_cost_map_model_name(completion_response_model, custom_llm_provider, priced_region),
-            requested=_cost_map_model_name(model, custom_llm_provider, None) if model is not None else None,
-            custom_llm_provider=custom_llm_provider,
-        )
+        return _served_or_priced_requested_model(completion_response_model, model, custom_llm_provider, priced_region)
 
     if return_model is None and model is not None:
         return_model = model
@@ -1035,8 +1034,14 @@ def _resolves_to_rates(model: str, custom_llm_provider: str | None) -> bool:
     return rule_info is not None and _cost_map_entry_prices_anything(rule_info)
 
 
-def _served_or_priced_requested_model(served: str, requested: str | None, custom_llm_provider: str | None) -> str:
-    if requested is None or requested == served or _resolves_to_rates(served, custom_llm_provider):
+def _served_or_priced_requested_model(
+    served_model: str, requested_model: str | None, custom_llm_provider: str | None, priced_region: str | None
+) -> str:
+    served: Final = _cost_map_model_name(served_model, custom_llm_provider, priced_region)
+    if requested_model is None:
+        return served
+    requested: Final = _cost_map_model_name(requested_model, custom_llm_provider, None)
+    if requested == served or _resolves_to_rates(served, custom_llm_provider):
         return served
     return requested if _resolves_to_rates(requested, custom_llm_provider) else served
 
diff --git a/tests/unit/test_cost_calculator.py b/tests/unit/test_cost_calculator.py
index 5886e0ce2ef..be6f178d674 100644
--- a/tests/unit/test_cost_calculator.py
+++ b/tests/unit/test_cost_calculator.py
@@ -456,6 +456,26 @@ def test_completion_cost_bills_requested_model_when_response_names_an_unmapped_d
     )
 
 
+def test_completion_cost_bills_requested_model_when_stream_reports_an_unmapped_display_name(
+    _local_model_cost_map: None, monkeypatch: pytest.MonkeyPatch
+) -> None:
+    requested_model: Final = "fireworks_ai/accounts/fireworks/models/sol-display-test"
+    monkeypatch.setitem(
+        litellm.model_cost,
+        requested_model,
+        {"litellm_provider": "fireworks_ai", "mode": "chat", "input_cost_per_token": 3e-6, "output_cost_per_token": 7e-6},
+    )
+    usage: Final = Usage(prompt_tokens=1000, completion_tokens=500, total_tokens=1500)
+    response: Final = ModelResponse(model="accounts/fireworks/models/sol-display-test", choices=[], usage=usage)
+    response._hidden_params["provider_response_model"] = "Sol Display Max"
+
+    cost: Final = completion_cost(
+        completion_response=response, model=requested_model, custom_llm_provider="fireworks_ai"
+    )
+
+    assert cost == pytest.approx(_token_rates_cost(litellm.model_cost[requested_model], usage))
+
+
 def test_completion_cost_bills_requested_model_when_response_row_has_no_rates(
     _local_model_cost_map: None, monkeypatch: pytest.MonkeyPatch
 ) -> None: