diff --git a/setup.cfg b/setup.cfg
index 46bc3b3ab1..c80073cedf 100644
--- a/setup.cfg
+++ b/setup.cfg
@@ -44,7 +44,7 @@ install_requires =
     openai>1  # For typing
     python-jose[cryptography]
     passlib[bcrypt]
-    aioprometheus[starlette]
+    aioprometheus[starlette]>=23.12.0
 
 [options.packages.find]
 exclude =
diff --git a/xinference/model/llm/llm_family.json b/xinference/model/llm/llm_family.json
index 768833c864..6d379ef87d 100644
--- a/xinference/model/llm/llm_family.json
+++ b/xinference/model/llm/llm_family.json
@@ -3184,7 +3184,7 @@
           "none"
         ],
         "model_id": "internlm/internlm2-chat-7b",
-        "model_revision": "5797f79825bab7013932d57e2babaac1b8de6b4f"
+        "model_revision": "2292b86b21cb856642782cebed0a453997453b1f"
       },
       {
         "model_format": "pytorch",
@@ -3193,22 +3193,22 @@
           "none"
         ],
         "model_id": "internlm/internlm2-chat-20b",
-        "model_revision": "3ccaf3ae82d5d01c0a95eecf40ee550f9c543635"
+        "model_revision": "b666125047cd98c5a7c85ca28720b44a06aed124"
       }
     ],
     "prompt_style": {
       "style_name": "INTERNLM2",
       "system_prompt": "You are InternLM (书生·浦语), a helpful, honest, and harmless AI assistant developed by Shanghai AI Laboratory (上海人工智能实验室).",
       "roles": [
-        "[UNUSED_TOKEN_146]user",
-        "[UNUSED_TOKEN_146]assistant"
+        "<|im_start|>user",
+        "<|im_start|>assistant"
       ],
-      "intra_message_sep": "[UNUSED_TOKEN_145]",
+      "intra_message_sep": "<|im_end|>",
       "stop_token_ids": [
         92542
       ],
       "stop": [
-        "[UNUSED_TOKEN_145]"
+        "<|im_end|>"
       ]
     }
   },
diff --git a/xinference/model/llm/llm_family_modelscope.json b/xinference/model/llm/llm_family_modelscope.json
index 0da91c6696..41161b95c6 100644
--- a/xinference/model/llm/llm_family_modelscope.json
+++ b/xinference/model/llm/llm_family_modelscope.json
@@ -1817,15 +1817,15 @@
       "style_name": "INTERNLM2",
       "system_prompt": "You are InternLM (书生·浦语), a helpful, honest, and harmless AI assistant developed by Shanghai AI Laboratory (上海人工智能实验室).",
       "roles": [
-        "[UNUSED_TOKEN_146]user",
-        "[UNUSED_TOKEN_146]assistant"
+        "<|im_start|>user",
+        "<|im_start|>assistant"
       ],
-      "intra_message_sep": "[UNUSED_TOKEN_145]",
+      "intra_message_sep": "<|im_end|>",
       "stop_token_ids": [
         92542
       ],
       "stop": [
-        "[UNUSED_TOKEN_145]"
+        "<|im_end|>"
       ]
     }
   },
diff --git a/xinference/model/llm/utils.py b/xinference/model/llm/utils.py
index 53ccbd7b9d..604bb780bf 100644
--- a/xinference/model/llm/utils.py
+++ b/xinference/model/llm/utils.py
@@ -310,7 +310,7 @@ def get_prompt(
             ret = (
                 "<s>"
                 if prompt_style.system_prompt == ""
-                else "<s>[UNUSED_TOKEN_146]system\n"
+                else "<s><|im_start|>system\n"
                 + prompt_style.system_prompt
                 + prompt_style.intra_message_sep
                 + "\n"