diff --git a/litellm/litellm_core_utils/litellm_logging.py b/litellm/litellm_core_utils/litellm_logging.py
index 5b4187846ff..4a27aac9769 100644
--- a/litellm/litellm_core_utils/litellm_logging.py
+++ b/litellm/litellm_core_utils/litellm_logging.py
@@ -640,8 +640,8 @@ class Logging(LiteLLMLoggingBaseClass):
         self._own_session_id: str = session_id_var.get()
 
         self.function_id = function_id
-        self.streaming_chunks: list[object] = []  # for generating complete stream response
-        self.sync_streaming_chunks: list[object] = []  # for generating complete stream response
+        self.streaming_chunks: list[object] = []
+        self.sync_streaming_chunks: list[object] = []
         self.log_raw_request_response = log_raw_request_response
         self.raw_request_only = raw_request_only
 
diff --git a/litellm/rust_bridge/public_call.py b/litellm/rust_bridge/public_call.py
index 17ff7e52165..d107483ecc0 100644
--- a/litellm/rust_bridge/public_call.py
+++ b/litellm/rust_bridge/public_call.py
@@ -6,6 +6,33 @@ import inspect
 from collections.abc import Callable, Mapping, Sequence
 from typing import Final, cast  # noqa: TID251  # narrows caller-owned containers without copying them
 
+import litellm
+
+_INFERENCE_CONTEXT: Final = frozenset(
+    {
+        "model",
+        "messages",
+        "input",
+        "api_key",
+        "api_base",
+        "base_url",
+        "custom_llm_provider",
+        "extra_headers",
+        "timeout",
+        "request_timeout",
+        "callbacks",
+        "success_callback",
+        "failure_callback",
+        "metadata",
+        "litellm_metadata",
+        "litellm_call_id",
+        "litellm_trace_id",
+        "litellm_logging_obj",
+        "litellm_credential_name",
+        "proxy_server_request",
+    }
+)
+
 
 def signature(legacy: Callable[..., object]) -> inspect.Signature:
     return inspect.signature(legacy)
@@ -43,37 +70,11 @@ def optional_sequence(value: object) -> Sequence[object] | None:
 
 
 def inference_decline_reason(parameters: tuple[str, ...], kwargs: Mapping[str, object]) -> str | None:
-    import litellm
-
     if litellm.cache is not None or litellm.drop_params or litellm.modify_params:
         return "native inference does not implement the configured cache or parameter rewrites"
-    context: Final = frozenset(
-        {
-            "model",
-            "messages",
-            "input",
-            "api_key",
-            "api_base",
-            "base_url",
-            "custom_llm_provider",
-            "extra_headers",
-            "timeout",
-            "request_timeout",
-            "callbacks",
-            "success_callback",
-            "failure_callback",
-            "metadata",
-            "litellm_metadata",
-            "litellm_call_id",
-            "litellm_trace_id",
-            "litellm_logging_obj",
-            "litellm_credential_name",
-            "proxy_server_request",
-        }
-    )
     for name, value in kwargs.items():
         if value is None:
             continue
-        if name not in parameters and name not in context:
+        if name not in parameters and name not in _INFERENCE_CONTEXT:
             return f"native inference does not implement {name}"
     return None
diff --git a/litellm/rust_bridge/responses/route_host.py b/litellm/rust_bridge/responses/route_host.py
index 110dd02ab22..4f491064185 100644
--- a/litellm/rust_bridge/responses/route_host.py
+++ b/litellm/rust_bridge/responses/route_host.py
@@ -34,7 +34,7 @@ def decline_reason(request: LiteLLMResponsesRequest) -> str | None:
     if request.custom_llm_provider is None and "/" not in request.model:
         try:
             _, provider, _, _ = get_llm_provider(model=request.model)
-        except Exception:  # noqa: BLE001  # unresolved models stay on the existing Python dispatch path
+        except litellm.exceptions.BadRequestError:
             return "native Responses could not resolve the provider"
         if provider != "openai":
             return "native HTTP responses provider"