diff --git a/application/llm/llama_cpp.py b/application/llm/llama_cpp.py
index 25a2f0c9..804c3c56 100644
--- a/application/llm/llama_cpp.py
+++ b/application/llm/llama_cpp.py
@@ -1,9 +1,30 @@
 from application.llm.base import BaseLLM
 from application.core.settings import settings
+import threading
+
+class LlamaSingleton:
+    _instances = {}
+    _lock = threading.Lock()  # Add a lock for thread synchronization
+
+    @classmethod
+    def get_instance(cls, llm_name):
+        if llm_name not in cls._instances:
+            try:
+                from llama_cpp import Llama
+            except ImportError:
+                raise ImportError(
+                    "Please install llama_cpp using pip install llama-cpp-python"
+                )
+            cls._instances[llm_name] = Llama(model_path=llm_name, n_ctx=2048)
+        return cls._instances[llm_name]
+
+    @classmethod
+    def query_model(cls, llm, prompt, **kwargs):
+        with cls._lock:
+            return llm(prompt, **kwargs)
 
 
 class LlamaCpp(BaseLLM):
-
     def __init__(
         self,
         api_key=None,
@@ -12,41 +33,23 @@ class LlamaCpp(BaseLLM):
         *args,
         **kwargs,
     ):
-        global llama
-        try:
-            from llama_cpp import Llama
-        except ImportError:
-            raise ImportError(
-                "Please install llama_cpp using pip install llama-cpp-python"
-            )
-
         super().__init__(*args, **kwargs)
         self.api_key = api_key
         self.user_api_key = user_api_key
-        llama = Llama(model_path=llm_name, n_ctx=2048)
+        self.llama = LlamaSingleton.get_instance(llm_name)
 
     def _raw_gen(self, baseself, model, messages, stream=False, **kwargs):
         context = messages[0]["content"]
         user_question = messages[-1]["content"]
         prompt = f"### Instruction \n {user_question} \n ### Context \n {context} \n ### Answer \n"
-
-        result = llama(prompt, max_tokens=150, echo=False)
-
-        # import sys
-        # print(result['choices'][0]['text'].split('### Answer \n')[-1], file=sys.stderr)
-
+        result = LlamaSingleton.query_model(self.llama, prompt, max_tokens=150, echo=False)
         return result["choices"][0]["text"].split("### Answer \n")[-1]
 
     def _raw_gen_stream(self, baseself, model, messages, stream=True, **kwargs):
         context = messages[0]["content"]
         user_question = messages[-1]["content"]
         prompt = f"### Instruction \n {user_question} \n ### Context \n {context} \n ### Answer \n"
-
-        result = llama(prompt, max_tokens=150, echo=False, stream=stream)
-
-        # import sys
-        # print(list(result), file=sys.stderr)
-
+        result = LlamaSingleton.query_model(self.llama, prompt, max_tokens=150, echo=False, stream=stream)
         for item in result:
             for choice in item["choices"]:
-                yield choice["text"]
+                yield choice["text"]
\ No newline at end of file