fix: constrain insufficient quota mapping

litellm/litellm_core_utils/exception_mapping_utils.py

@@ -250,6 +250,7 @@ class _ProviderHTTPException(Protocol):
 
 @runtime_checkable
 class _ProviderQuotaException(Protocol):
+    status_code: int
     body: object
     response: httpx.Response
 
@@ -257,6 +258,8 @@ class _ProviderQuotaException(Protocol):
 def _get_insufficient_quota_response(original_exception: object) -> Optional[httpx.Response]:
     if not isinstance(original_exception, _ProviderQuotaException):
         return None
+    if original_exception.status_code != 429:
+        return None
     body = original_exception.body
     if not isinstance(body, Mapping):
         return None
@@ -301,7 +304,7 @@ def _map_openai_exception(
     insufficient_quota_response = _get_insufficient_quota_response(original_exception)
     if insufficient_quota_response is not None:
         raise InsufficientQuotaError(
-            message=f"InsufficientQuotaError: {exception_provider} - {message}",
+            message=f"{exception_provider} - {message}",
             model=model,
             llm_provider=custom_llm_provider,
             response=insufficient_quota_response,

tests/test_litellm/litellm_core_utils/test_exception_mapping_utils.py

@@ -230,6 +230,7 @@ def test_openai_insufficient_quota_maps_to_distinct_rate_limit_subtype(body):
     assert isinstance(exc_info.value, litellm.RateLimitError)
     assert exc_info.value.code == "insufficient_quota"
     assert exc_info.value.type == "insufficient_quota"
+    assert str(exc_info.value).count("InsufficientQuotaError") == 1
     assert litellm.InsufficientQuotaError in litellm.LITELLM_EXCEPTION_TYPES
 
 
@@ -265,6 +266,24 @@ def test_openai_transient_429_remains_rate_limit_error(body):
     assert exc_info.value.type == "throttling_error"
 
 
+def test_openai_non_429_insufficient_quota_body_is_not_quota_error():
+    original_exception = OpenAIError(
+        status_code=403,
+        message="Forbidden",
+        headers={},
+        body={"error": {"code": "insufficient_quota"}},
+    )
+
+    with pytest.raises(litellm.APIError) as exc_info:
+        exception_type(
+            model="gpt-5.5",
+            original_exception=original_exception,
+            custom_llm_provider="openai",
+        )
+
+    assert type(exc_info.value) is litellm.APIError
+
+
 gemini_context_window_test_cases = [
     # Gemini 2.0 Flash format (includes input token count in message)
     (