diff --git a/python/sglang/srt/entrypoints/openai/protocol.py b/python/sglang/srt/entrypoints/openai/protocol.py
index b0825dfba342..6e46ac6a31b1 100644
--- a/python/sglang/srt/entrypoints/openai/protocol.py
+++ b/python/sglang/srt/entrypoints/openai/protocol.py
@@ -475,7 +475,7 @@ class CompletionResponseChoice(BaseModel):
     finish_reason: Optional[Literal["stop", "length", "content_filter", "abort"]] = None
     matched_stop: Union[None, int, str] = None
     hidden_states: Optional[object] = None
-    token_ids: Optional[List[int]] = None
+    response_token_ids: Optional[List[int]] = None
     prompt_token_ids: Optional[List[int]] = None
 
     @model_serializer(mode="wrap")
@@ -483,8 +483,8 @@ def _serialize(self, handler):
         data = handler(self)
         if self.hidden_states is None:
             data.pop("hidden_states", None)
-        if self.token_ids is None:
-            data.pop("token_ids", None)
+        if self.response_token_ids is None:
+            data.pop("response_token_ids", None)
         if self.prompt_token_ids is None:
             data.pop("prompt_token_ids", None)
         return data
@@ -515,7 +515,7 @@ class CompletionResponseStreamChoice(BaseModel):
     finish_reason: Optional[Literal["stop", "length", "content_filter", "abort"]] = None
     matched_stop: Union[None, int, str] = None
     hidden_states: Optional[object] = None
-    token_ids: Optional[List[int]] = None
+    response_token_ids: Optional[List[int]] = None
     prompt_token_ids: Optional[List[int]] = None
 
     @model_serializer(mode="wrap")
@@ -523,8 +523,8 @@ def _serialize(self, handler):
         data = handler(self)
         if self.hidden_states is None:
             data.pop("hidden_states", None)
-        if self.token_ids is None:
-            data.pop("token_ids", None)
+        if self.response_token_ids is None:
+            data.pop("response_token_ids", None)
         if self.prompt_token_ids is None:
             data.pop("prompt_token_ids", None)
         return data
diff --git a/python/sglang/srt/entrypoints/openai/serving_completions.py b/python/sglang/srt/entrypoints/openai/serving_completions.py
index b7da6fdf2b7a..6c797d2be460 100644
--- a/python/sglang/srt/entrypoints/openai/serving_completions.py
+++ b/python/sglang/srt/entrypoints/openai/serving_completions.py
@@ -374,7 +374,7 @@ async def _generate_completion_stream(
                         if finish_reason and "matched" in finish_reason
                         else None
                     ),
-                    token_ids=chunk_token_ids,
+                    response_token_ids=chunk_token_ids,
                     prompt_token_ids=chunk_prompt_token_ids,
                 )
                 chunk = CompletionStreamResponse(
@@ -609,7 +609,7 @@ def _build_completion_response(
                     else None
                 ),
                 hidden_states=hidden_states,
-                token_ids=(
+                response_token_ids=(
                     ret_item["output_ids"] if request.return_token_ids else None
                 ),
                 prompt_token_ids=(
diff --git a/test/registered/unit/entrypoints/openai/test_serving_completions.py b/test/registered/unit/entrypoints/openai/test_serving_completions.py
index 289d09ec2e43..9f1f3d549157 100644
--- a/test/registered/unit/entrypoints/openai/test_serving_completions.py
+++ b/test/registered/unit/entrypoints/openai/test_serving_completions.py
@@ -249,8 +249,11 @@ def test_non_streaming_response(self):
         self.assertEqual(len(response.choices), 1)
         self.assertEqual(response.choices[0].text, " world")
         self.assertEqual(len(response.choices[0].logprobs.top_logprobs), 0)
-        self.assertEqual(response.choices[0].token_ids, [3, 4])
+        self.assertEqual(response.choices[0].response_token_ids, [3, 4])
         self.assertEqual(response.choices[0].prompt_token_ids, [1, 2])
+        dumped_choice = response.model_dump()["choices"][0]
+        self.assertNotIn("token_ids", dumped_choice)
+        self.assertEqual(dumped_choice["response_token_ids"], [3, 4])
 
     def test_streaming_abort_yields_error(self):
         """Test that an abort finish reason during streaming correctly yields an error and stops."""
@@ -383,10 +386,14 @@ async def run_stream():
                     data = json.loads(raw[len("data: ") :])
                     choices.extend(data.get("choices", []))
 
-                token_ids = [tid for c in choices for tid in c.get("token_ids", [])]
+                token_ids = [
+                    tid for c in choices for tid in c.get("response_token_ids", [])
+                ]
                 text = "".join(c["text"] for c in choices)
                 self.assertEqual(text, "abc")
                 self.assertEqual(token_ids, [5, 6, 7])
+                for choice in choices:
+                    self.assertNotIn("token_ids", choice)
                 self.assertEqual(choices[0]["prompt_token_ids"], [1, 2])
                 for choice in choices[1:]:
                     self.assertNotIn("prompt_token_ids", choice)