diff --git a/litellm/litellm_core_utils/llm_response_utils/convert_dict_to_response.py b/litellm/litellm_core_utils/llm_response_utils/convert_dict_to_response.py
index 8b94c34942c..6dee99375ac 100644
--- a/litellm/litellm_core_utils/llm_response_utils/convert_dict_to_response.py
+++ b/litellm/litellm_core_utils/llm_response_utils/convert_dict_to_response.py
@@ -422,6 +422,7 @@ class _PromptCacheTokenCounts(TypedDict, total=False):
 _CACHED_MESSAGE_PROVIDER_FIELDS: Final = TypeAdapter(_CachedMessageProviderFields)
 _CACHED_USAGE_DETAILS: Final = TypeAdapter(_CachedUsageDetails)
 _FIELD_MAP: Final = TypeAdapter(dict[str, object])
+_CHOICE_FIELD_MAPS: Final = TypeAdapter(list[dict[str, object]])
 _COMPLETION_TOKENS_DETAILS: Final = TypeAdapter(CompletionTokensDetailsWrapper)
 _PROMPT_TOKENS_DETAILS: Final = TypeAdapter(PromptTokensDetailsWrapper)
 _PROMPT_CACHE_TOKEN_COUNTS: Final = TypeAdapter(_PromptCacheTokenCounts)
@@ -757,6 +758,20 @@ def convert_to_model_response_object(
                     model="",
                 )
 
+            response_fields: Final = _FIELD_MAP.validate_python(response_object)
+            if any(
+                choice.get("finish_reason") == "abort"
+                for choice in _CHOICE_FIELD_MAPS.validate_python(response_fields["choices"])
+            ):
+                from litellm.exceptions import APIError
+
+                raise APIError(
+                    status_code=500,
+                    message="Upstream generation aborted (finish_reason=abort)",
+                    llm_provider="",
+                    model=str(response_fields.get("model") or ""),
+                )
+
             for idx, choice in enumerate(response_object["choices"]):
                 ## HANDLE JSON MODE - anthropic returns single function call]
                 tool_calls = choice["message"].get("tool_calls", None)
diff --git a/tests/unit/litellm_core_utils/llm_response_utils/test_convert_dict_to_response.py b/tests/unit/litellm_core_utils/llm_response_utils/test_convert_dict_to_response.py
index 0c601a6b2c8..e21a2c9e296 100644
--- a/tests/unit/litellm_core_utils/llm_response_utils/test_convert_dict_to_response.py
+++ b/tests/unit/litellm_core_utils/llm_response_utils/test_convert_dict_to_response.py
@@ -228,6 +228,30 @@ def test_convert_missing_choices_raises_api_error() -> None:
     assert "no 'choices'" in str(exc_info.value)
 
 
+@pytest.mark.parametrize("finish_reasons", [("abort",), ("stop", "abort"), ("abort", "stop")])
+def test_convert_aborted_choice_raises_api_error(finish_reasons: tuple[str, ...]) -> None:
+    from litellm.exceptions import APIError
+
+    resp: Final = {
+        "id": "x",
+        "created": 1,
+        "model": "Qwen/Qwen3-8B",
+        "object": "chat.completion",
+        "choices": [
+            {"index": index, "message": {"role": "assistant", "content": "Oslo is 4C and"}, "finish_reason": reason}
+            for index, reason in enumerate(finish_reasons)
+        ],
+    }
+    with pytest.raises(APIError, match="finish_reason=abort") as exc_info:
+        convert_to_model_response_object(
+            response_object=resp,
+            model_response_object=ModelResponse(),
+            response_type="completion",
+        )
+    assert exc_info.value.status_code == 500
+    assert exc_info.value.model == "Qwen/Qwen3-8B"
+
+
 @pytest.mark.parametrize(("choices", "type_name"), [({}, "dict"), ("", "str"), (None, "NoneType"), (0, "int")])
 @pytest.mark.asyncio
 async def test_convert_non_list_choices_raises_api_error(choices: object, type_name: str) -> None: