From 239bcbc21423ee09bc8e58966e9d34a29a5bb6cb Mon Sep 17 00:00:00 2001
From: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 00:44:30 +0000
Subject: [PATCH 1/5] fix(fireworks): resolve short model names to long cost
 map keys

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---
 litellm/utils.py                 |  5 +++
 tests/test_litellm/test_utils.py | 69 ++++++++++++++++++++++++++++++++
 2 files changed, 74 insertions(+)

diff --git a/litellm/utils.py b/litellm/utils.py
index 394ab4b4094..4a92f9b1244 100644
--- a/litellm/utils.py
+++ b/litellm/utils.py
@@ -5556,6 +5556,11 @@ def _get_potential_model_names(model: str, custom_llm_provider: str | None) -> P
 
         split_model = strip_bedrock_routing_prefix(split_model)
 
+    if custom_llm_provider == "fireworks_ai":
+        from litellm.llms.fireworks_ai.common_utils import resolve_fireworks_resource_name
+
+        provider_prefixed_model_name = f"fireworks_ai/{resolve_fireworks_resource_name(split_model)}"
+
     return PotentialModelNamesAndCustomLLMProvider(
         split_model=split_model,
         combined_model_name=combined_model_name,
diff --git a/tests/test_litellm/test_utils.py b/tests/test_litellm/test_utils.py
index 835e87aff88..02e2eeec467 100644
--- a/tests/test_litellm/test_utils.py
+++ b/tests/test_litellm/test_utils.py
@@ -4448,6 +4448,75 @@ def test_fireworks_models_in_backup_cost_map():
         ), f"short-form {short_key} does not match long-form {long_key}"
 
 
+@pytest.fixture
+def fireworks_short_model_cost_map(monkeypatch: pytest.MonkeyPatch) -> Iterator[None]:
+    monkeypatch.setattr(
+        litellm,
+        "model_cost",
+        {
+            "fireworks_ai/accounts/fireworks/models/glm-5p3": {
+                "input_cost_per_token": 1e-6,
+                "output_cost_per_token": 2e-6,
+                "litellm_provider": "fireworks_ai",
+                "mode": "chat",
+                "max_tokens": 100,
+            },
+            "fireworks_ai/accounts/fireworks/routers/glm-5p3-fast": {
+                "input_cost_per_token": 2.1e-6,
+                "output_cost_per_token": 6.6e-6,
+                "litellm_provider": "fireworks_ai",
+                "mode": "chat",
+            },
+            "fireworks_ai/nomic-ai/nomic-embed-text-v1.5": {
+                "input_cost_per_token": 8e-9,
+                "output_cost_per_token": 0.0,
+                "litellm_provider": "fireworks_ai",
+                "mode": "embedding",
+            },
+        },
+    )
+    litellm.get_model_info.cache_clear()
+    yield
+    litellm.get_model_info.cache_clear()
+
+
+def test_fireworks_short_model_names_resolve_to_long_cost_map_keys(fireworks_short_model_cost_map: None) -> None:
+    model_info = litellm.get_model_info("fireworks_ai/glm-5p3")
+    assert model_info["key"] == "fireworks_ai/accounts/fireworks/models/glm-5p3"
+    assert model_info["input_cost_per_token"] == 1e-6
+    assert model_info["max_tokens"] == 100
+
+    model_info = litellm.get_model_info("glm-5p3", custom_llm_provider="fireworks_ai")
+    assert model_info["key"] == "fireworks_ai/accounts/fireworks/models/glm-5p3"
+
+    model_info = litellm.get_model_info("fireworks_ai/glm-5p3-fast")
+    assert model_info["key"] == "fireworks_ai/accounts/fireworks/routers/glm-5p3-fast"
+    assert model_info["input_cost_per_token"] == 2.1e-6
+
+    model_info = litellm.get_model_info("fireworks_ai/nomic-ai/nomic-embed-text-v1.5")
+    assert model_info["key"] == "fireworks_ai/nomic-ai/nomic-embed-text-v1.5"
+
+    with pytest.raises(Exception, match="isn't mapped"):
+        litellm.get_model_info("fireworks_ai/does-not-exist")
+
+
+def test_fireworks_short_model_names_price_with_completion_cost(fireworks_short_model_cost_map: None) -> None:
+    from litellm.types.utils import ModelResponse
+
+    response = ModelResponse(
+        model="fireworks_ai/glm-5p3",
+        usage=Usage(prompt_tokens=10, completion_tokens=5, total_tokens=15),
+    )
+
+    cost = litellm.completion_cost(
+        completion_response=response,
+        model="fireworks_ai/glm-5p3",
+        custom_llm_provider="fireworks_ai",
+    )
+
+    assert cost == pytest.approx(10 * 1e-6 + 5 * 2e-6)
+
+
 class TestBedrockBaseModelLabelKeepsTools:
     """Regression for #29618: a Bedrock deployment whose ``base_model`` is a friendly
     label must not silently drop ``tools``/``tool_choice`` under ``drop_params``."""

From 0904051fda4e3f07c59ecca731eb3c5b94b92817 Mon Sep 17 00:00:00 2001
From: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 03:24:23 +0000
Subject: [PATCH 2/5] refactor(fireworks): move cost map key construction under
 llms/

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---
 litellm/llms/fireworks_ai/common_utils.py        |  4 ++++
 litellm/utils.py                                 |  4 ++--
 .../test_fireworks_ai_common_utils.py            | 16 +++++++++++++---
 3 files changed, 19 insertions(+), 5 deletions(-)

diff --git a/litellm/llms/fireworks_ai/common_utils.py b/litellm/llms/fireworks_ai/common_utils.py
index 21a630a76d7..b5552583862 100644
--- a/litellm/llms/fireworks_ai/common_utils.py
+++ b/litellm/llms/fireworks_ai/common_utils.py
@@ -72,6 +72,10 @@ def resolve_fireworks_resource_name(model: str) -> str:
     return f"accounts/fireworks/models/{stripped}"
 
 
+def fireworks_cost_map_key(model: str) -> str:
+    return f"fireworks_ai/{resolve_fireworks_resource_name(model)}"
+
+
 class FireworksAIMixin:
     """
     Common Base Config functions across Fireworks AI Endpoints
diff --git a/litellm/utils.py b/litellm/utils.py
index 4a92f9b1244..d088ae7264f 100644
--- a/litellm/utils.py
+++ b/litellm/utils.py
@@ -5557,9 +5557,9 @@ def _get_potential_model_names(model: str, custom_llm_provider: str | None) -> P
         split_model = strip_bedrock_routing_prefix(split_model)
 
     if custom_llm_provider == "fireworks_ai":
-        from litellm.llms.fireworks_ai.common_utils import resolve_fireworks_resource_name
+        from litellm.llms.fireworks_ai.common_utils import fireworks_cost_map_key
 
-        provider_prefixed_model_name = f"fireworks_ai/{resolve_fireworks_resource_name(split_model)}"
+        provider_prefixed_model_name = fireworks_cost_map_key(split_model)
 
     return PotentialModelNamesAndCustomLLMProvider(
         split_model=split_model,
diff --git a/tests/test_litellm/llms/fireworks_ai/test_fireworks_ai_common_utils.py b/tests/test_litellm/llms/fireworks_ai/test_fireworks_ai_common_utils.py
index 16226a3ce74..5aa1429f2e7 100644
--- a/tests/test_litellm/llms/fireworks_ai/test_fireworks_ai_common_utils.py
+++ b/tests/test_litellm/llms/fireworks_ai/test_fireworks_ai_common_utils.py
@@ -1,8 +1,6 @@
-
 import pytest
 
-
-from litellm.llms.fireworks_ai.common_utils import resolve_fireworks_resource_name
+from litellm.llms.fireworks_ai.common_utils import fireworks_cost_map_key, resolve_fireworks_resource_name
 
 
 @pytest.mark.parametrize(
@@ -43,3 +41,15 @@ from litellm.llms.fireworks_ai.common_utils import resolve_fireworks_resource_na
 )
 def test_resolve_fireworks_resource_name(model, expected):
     assert resolve_fireworks_resource_name(model) == expected
+
+
+@pytest.mark.parametrize(
+    "model, expected",
+    [
+        ("deepseek-r1", "fireworks_ai/accounts/fireworks/models/deepseek-r1"),
+        ("glm-5p3-fast", "fireworks_ai/accounts/fireworks/routers/glm-5p3-fast"),
+        ("accounts/fireworks/models/deepseek-r1", "fireworks_ai/accounts/fireworks/models/deepseek-r1"),
+    ],
+)
+def test_fireworks_cost_map_key(model: str, expected: str) -> None:
+    assert fireworks_cost_map_key(model) == expected

From a519d805bb35abad5ba736d0e785aebe813cca26 Mon Sep 17 00:00:00 2001
From: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 03:29:11 +0000
Subject: [PATCH 3/5] refactor(fireworks): resolve cost map key through a
 provider config hook

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---
 litellm/llms/base_llm/base_utils.py              |  6 ++++++
 litellm/llms/fireworks_ai/chat/transformation.py |  3 +++
 litellm/llms/fireworks_ai/common_utils.py        |  4 ----
 litellm/utils.py                                 | 14 +++++++++-----
 .../test_fireworks_ai_chat_transformation.py     | 12 ++++++++++++
 .../test_fireworks_ai_common_utils.py            | 16 +++-------------
 6 files changed, 33 insertions(+), 22 deletions(-)

diff --git a/litellm/llms/base_llm/base_utils.py b/litellm/llms/base_llm/base_utils.py
index c5290b41f7b..c9ef7d77ed8 100644
--- a/litellm/llms/base_llm/base_utils.py
+++ b/litellm/llms/base_llm/base_utils.py
@@ -50,6 +50,12 @@ class BaseLLMModelInfo(ABC):
         """
         return None
 
+    def get_model_cost_key(self, model: str) -> str | None:
+        """
+        Extra `litellm.model_cost` key to try for this provider's spelling of `model`, after the exact keys miss.
+        """
+        return None
+
     @abstractmethod
     def get_models(self, api_key: str | None = None, api_base: str | None = None) -> list[str]:
         """
diff --git a/litellm/llms/fireworks_ai/chat/transformation.py b/litellm/llms/fireworks_ai/chat/transformation.py
index b4e1856f499..05160d83c12 100644
--- a/litellm/llms/fireworks_ai/chat/transformation.py
+++ b/litellm/llms/fireworks_ai/chat/transformation.py
@@ -602,6 +602,9 @@ class FireworksAIConfig(FireworksAIMixin, OpenAIGPTConfig):
             return None
         return max(matches, key=lambda match: len(match[0]))[1]
 
+    def get_model_cost_key(self, model: str) -> str:
+        return f"fireworks_ai/{resolve_fireworks_resource_name(model)}"
+
     def get_provider_info(self, model: str) -> ProviderSpecificModelInfo:
         supports_function_calling_value: Final = self._get_model_cost_capability(
             model=model, capability="supports_function_calling"
diff --git a/litellm/llms/fireworks_ai/common_utils.py b/litellm/llms/fireworks_ai/common_utils.py
index b5552583862..21a630a76d7 100644
--- a/litellm/llms/fireworks_ai/common_utils.py
+++ b/litellm/llms/fireworks_ai/common_utils.py
@@ -72,10 +72,6 @@ def resolve_fireworks_resource_name(model: str) -> str:
     return f"accounts/fireworks/models/{stripped}"
 
 
-def fireworks_cost_map_key(model: str) -> str:
-    return f"fireworks_ai/{resolve_fireworks_resource_name(model)}"
-
-
 class FireworksAIMixin:
     """
     Common Base Config functions across Fireworks AI Endpoints
diff --git a/litellm/utils.py b/litellm/utils.py
index d088ae7264f..edfdde55699 100644
--- a/litellm/utils.py
+++ b/litellm/utils.py
@@ -5556,17 +5556,21 @@ def _get_potential_model_names(model: str, custom_llm_provider: str | None) -> P
 
         split_model = strip_bedrock_routing_prefix(split_model)
 
-    if custom_llm_provider == "fireworks_ai":
-        from litellm.llms.fireworks_ai.common_utils import fireworks_cost_map_key
-
-        provider_prefixed_model_name = fireworks_cost_map_key(split_model)
+    provider_model_info: Final = (
+        ProviderConfigManager.get_provider_model_info(model=split_model, provider=LlmProviders(custom_llm_provider))
+        if custom_llm_provider in LlmProvidersSet
+        else None
+    )
+    provider_cost_key: Final = (
+        provider_model_info.get_model_cost_key(split_model) if provider_model_info is not None else None
+    )
 
     return PotentialModelNamesAndCustomLLMProvider(
         split_model=split_model,
         combined_model_name=combined_model_name,
         stripped_model_name=stripped_model_name,
         combined_stripped_model_name=combined_stripped_model_name,
-        provider_prefixed_model_name=provider_prefixed_model_name,
+        provider_prefixed_model_name=provider_cost_key or provider_prefixed_model_name,
         custom_llm_provider=cast(str, custom_llm_provider),
     )
 
diff --git a/tests/test_litellm/llms/fireworks_ai/chat/test_fireworks_ai_chat_transformation.py b/tests/test_litellm/llms/fireworks_ai/chat/test_fireworks_ai_chat_transformation.py
index 8479397efa7..fb0311ef39b 100644
--- a/tests/test_litellm/llms/fireworks_ai/chat/test_fireworks_ai_chat_transformation.py
+++ b/tests/test_litellm/llms/fireworks_ai/chat/test_fireworks_ai_chat_transformation.py
@@ -1813,3 +1813,15 @@ def test_streaming_preserves_selected_model_for_private_accounting():
         completion_response=assembled,
         custom_llm_provider="fireworks_ai",
     ) == pytest.approx(expected_cost)
+
+
+@pytest.mark.parametrize(
+    "model, expected",
+    [
+        ("deepseek-r1", "fireworks_ai/accounts/fireworks/models/deepseek-r1"),
+        ("glm-5p3-fast", "fireworks_ai/accounts/fireworks/routers/glm-5p3-fast"),
+        ("accounts/fireworks/models/deepseek-r1", "fireworks_ai/accounts/fireworks/models/deepseek-r1"),
+    ],
+)
+def test_get_model_cost_key_resolves_short_names_to_long_keys(model: str, expected: str) -> None:
+    assert FireworksAIConfig().get_model_cost_key(model) == expected
diff --git a/tests/test_litellm/llms/fireworks_ai/test_fireworks_ai_common_utils.py b/tests/test_litellm/llms/fireworks_ai/test_fireworks_ai_common_utils.py
index 5aa1429f2e7..16226a3ce74 100644
--- a/tests/test_litellm/llms/fireworks_ai/test_fireworks_ai_common_utils.py
+++ b/tests/test_litellm/llms/fireworks_ai/test_fireworks_ai_common_utils.py
@@ -1,6 +1,8 @@
+
 import pytest
 
-from litellm.llms.fireworks_ai.common_utils import fireworks_cost_map_key, resolve_fireworks_resource_name
+
+from litellm.llms.fireworks_ai.common_utils import resolve_fireworks_resource_name
 
 
 @pytest.mark.parametrize(
@@ -41,15 +43,3 @@ from litellm.llms.fireworks_ai.common_utils import fireworks_cost_map_key, resol
 )
 def test_resolve_fireworks_resource_name(model, expected):
     assert resolve_fireworks_resource_name(model) == expected
-
-
-@pytest.mark.parametrize(
-    "model, expected",
-    [
-        ("deepseek-r1", "fireworks_ai/accounts/fireworks/models/deepseek-r1"),
-        ("glm-5p3-fast", "fireworks_ai/accounts/fireworks/routers/glm-5p3-fast"),
-        ("accounts/fireworks/models/deepseek-r1", "fireworks_ai/accounts/fireworks/models/deepseek-r1"),
-    ],
-)
-def test_fireworks_cost_map_key(model: str, expected: str) -> None:
-    assert fireworks_cost_map_key(model) == expected

From f1fe61af06ddefcf086ec831cc172107c6e8c886 Mon Sep 17 00:00:00 2001
From: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 03:36:00 +0000
Subject: [PATCH 4/5] docs(fireworks): explain what get_model_cost_key is for

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---
 litellm/llms/base_llm/base_utils.py | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/litellm/llms/base_llm/base_utils.py b/litellm/llms/base_llm/base_utils.py
index c9ef7d77ed8..32f98fbc933 100644
--- a/litellm/llms/base_llm/base_utils.py
+++ b/litellm/llms/base_llm/base_utils.py
@@ -52,7 +52,10 @@ class BaseLLMModelInfo(ABC):
 
     def get_model_cost_key(self, model: str) -> str | None:
         """
-        Extra `litellm.model_cost` key to try for this provider's spelling of `model`, after the exact keys miss.
+        Maps the model name a user sends to the key `litellm.model_cost` stores it under, when the two differ.
+        `get_model_info` tries this key once the exact `model` and `provider/model` keys miss. The default None means
+        the provider's user-facing names already match the cost map, so there is nothing extra to try. Fireworks
+        overrides it: `deepseek-r1` -> `fireworks_ai/accounts/fireworks/models/deepseek-r1`.
         """
         return None
 

From 7a0ea9867df1573f3847b6cd227dd39be7c2a8fa Mon Sep 17 00:00:00 2001
From: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Date: Sun, 13 Sep 2026 03:38:14 +0000
Subject: [PATCH 5/5] docs(fireworks): drop provider example from
 get_model_cost_key docstring

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---
 litellm/llms/base_llm/base_utils.py | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/litellm/llms/base_llm/base_utils.py b/litellm/llms/base_llm/base_utils.py
index 32f98fbc933..a446a721307 100644
--- a/litellm/llms/base_llm/base_utils.py
+++ b/litellm/llms/base_llm/base_utils.py
@@ -54,8 +54,7 @@ class BaseLLMModelInfo(ABC):
         """
         Maps the model name a user sends to the key `litellm.model_cost` stores it under, when the two differ.
         `get_model_info` tries this key once the exact `model` and `provider/model` keys miss. The default None means
-        the provider's user-facing names already match the cost map, so there is nothing extra to try. Fireworks
-        overrides it: `deepseek-r1` -> `fireworks_ai/accounts/fireworks/models/deepseek-r1`.
+        the provider's user-facing names already match the cost map, so there is nothing extra to try.
         """
         return None