diff --git a/vllm/entrypoints/openai/responses/serving.py b/vllm/entrypoints/openai/responses/serving.py
index 6d0041813e35..45e1dcbca30c 100644
--- a/vllm/entrypoints/openai/responses/serving.py
+++ b/vllm/entrypoints/openai/responses/serving.py
@@ -903,6 +903,7 @@ def _make_response_output_items(
             parser = self.parser(tokenizer)
             return parser.extract_response_outputs(
                 model_output=final_output.text,
+                model_output_token_ids=final_output.token_ids,
                 request=request,
                 enable_auto_tools=self.enable_auto_tools,
                 tool_call_id_type=self.tool_call_id_type,
diff --git a/vllm/parser/abstract_parser.py b/vllm/parser/abstract_parser.py
index aa145bab2121..51ea5e57c9db 100644
--- a/vllm/parser/abstract_parser.py
+++ b/vllm/parser/abstract_parser.py
@@ -154,7 +154,9 @@ def extract_content_ids(self, input_ids: list[int]) -> list[int]:
     @abstractmethod
     def extract_response_outputs(
         self,
+        *,
         model_output: str,
+        model_output_token_ids: Sequence[int],
         request: ResponsesRequest,
         enable_auto_tools: bool = False,
         tool_call_id_type: str = "random",
@@ -169,6 +171,7 @@ def extract_response_outputs(
 
         Args:
             model_output: The complete model-generated string.
+            model_output_token_ids: The token IDs of the model output.
             request: The request object used to generate the output.
             enable_auto_tools: Whether to enable automatic tool call parsing.
             tool_call_id_type: Type of tool call ID generation ("random", etc).
@@ -312,7 +315,9 @@ def extract_reasoning(
 
     def extract_response_outputs(
         self,
+        *,
         model_output: str,
+        model_output_token_ids: Sequence[int],
         request: ResponsesRequest,
         enable_auto_tools: bool = False,
         tool_call_id_type: str = "random",