diff --git a/fastdeploy/input/qwen_vl_processor/qwen_vl_processor.py b/fastdeploy/input/qwen_vl_processor/qwen_vl_processor.py
index dc85b78c04c..a3adeddf122 100644
--- a/fastdeploy/input/qwen_vl_processor/qwen_vl_processor.py
+++ b/fastdeploy/input/qwen_vl_processor/qwen_vl_processor.py
@@ -69,7 +69,7 @@ def __init__(
             tokenizer=self.tokenizer,
             **processor_kwargs,
         )
-
+        self.image_patch_id = self.processor.image_token_id
         self.limit_mm_per_prompt = self._parse_limits(limit_mm_per_prompt)
 
     def process_request(self, request, max_model_len=None, **kwargs):
@@ -249,6 +249,16 @@ def process_request_dict(self, request, max_model_len=None):
         # Handle continuation of previous generation by appending existing tokens
         if metadata and metadata.get("generated_token_ids"):
             self.append_generated_tokens(outputs, metadata["generated_token_ids"])
+
+        enable_thinking = False
+        if metadata:
+            enable_thinking = metadata.get("enable_thinking", False)
+
+        if request.get("chat_template_kwargs"):
+            chat_template_kwargs = request.get("chat_template_kwargs")
+            enable_thinking = chat_template_kwargs.get("enable_thinking", False)
+        request["enable_thinking"] = enable_thinking
+
         outputs = self.pack_outputs(outputs)
 
         request["prompt_token_ids"] = outputs["input_ids"].tolist()