diff --git a/vllm/entrypoints/openai/protocol.py b/vllm/entrypoints/openai/protocol.py
index fc15b7833ecf..ea879a019102 100644
--- a/vllm/entrypoints/openai/protocol.py
+++ b/vllm/entrypoints/openai/protocol.py
@@ -78,6 +78,8 @@ class ChatCompletionRequest(BaseModel):
     echo: Optional[bool] = False
     repetition_penalty: Optional[float] = 1.0
     min_p: Optional[float] = 0.0
+    prefix_pos: Optional[int] = None
+    prefix_stop: Optional[str] = None
     include_stop_str_in_output: Optional[bool] = False
     length_penalty: Optional[float] = 1.0
 
@@ -131,6 +133,8 @@ class CompletionRequest(BaseModel):
     spaces_between_special_tokens: Optional[bool] = True
     repetition_penalty: Optional[float] = 1.0
     min_p: Optional[float] = 0.0
+    prefix_pos: Optional[int] = None
+    prefix_stop: Optional[str] = None
     include_stop_str_in_output: Optional[bool] = False
     length_penalty: Optional[float] = 1.0
 
diff --git a/vllm/entrypoints/openai/serving_chat.py b/vllm/entrypoints/openai/serving_chat.py
index a9e4c355560b..c39acb5b8e1d 100644
--- a/vllm/entrypoints/openai/serving_chat.py
+++ b/vllm/entrypoints/openai/serving_chat.py
@@ -67,8 +67,16 @@ async def create_chat_completion(
         except ValueError as e:
             return self.create_error_response(str(e))
 
+        prefix_pos = request.prefix_pos
+        if request.prefix_stop is not None and request.prefix_stop in prompt:
+            prefix_index = prompt.index(request.prefix_stop)
+            prefix_pos = len(self.tokenizer.encode(prompt[:prefix_index])) - 1
+            prompt = prompt.replace(request.prefix_stop, '')
+
         result_generator = self.engine.generate(prompt, sampling_params,
-                                                request_id, token_ids)
+                                                request_id, token_ids,
+                                                prefix_pos)
+
         # Streaming response
         if request.stream:
             return self.chat_completion_stream_generator(
diff --git a/vllm/entrypoints/openai/serving_completion.py b/vllm/entrypoints/openai/serving_completion.py
index 8c9a7ad309ce..1b48543a72c1 100644
--- a/vllm/entrypoints/openai/serving_completion.py
+++ b/vllm/entrypoints/openai/serving_completion.py
@@ -286,11 +286,19 @@ async def create_completion(self, request: CompletionRequest,
             sampling_params = request.to_sampling_params()
             prompt_is_tokens, prompts = parse_prompt_format(request.prompt)
 
+            prefix_pos = request.prefix_pos
             for i, prompt in enumerate(prompts):
                 if prompt_is_tokens:
                     input_ids = self._validate_prompt_and_tokenize(
                         request, prompt_ids=prompt)
                 else:
+                    # Parse prefix position by prompt and prefix_stop indicator
+                    if request.prefix_stop is not None and request.prefix_stop in prompt:
+                        prefix_index = prompt.index(request.prefix_stop)
+                        prefix_pos = len(
+                            self.tokenizer.encode(prompt[:prefix_index])) - 1
+                        prompt = prompt.replace(request.prefix_stop, '')
+
                     input_ids = self._validate_prompt_and_tokenize(
                         request, prompt=prompt)
 
@@ -298,7 +306,8 @@ async def create_completion(self, request: CompletionRequest,
                     self.engine.generate(None,
                                          sampling_params,
                                          f"{request_id}-{i}",
-                                         prompt_token_ids=input_ids))
+                                         prompt_token_ids=input_ids,
+                                         prefix_pos=prefix_pos))
         except ValueError as e:
             return self.create_error_response(str(e))