diff --git a/python/paddle_serving_server_gpu/__init__.py b/python/paddle_serving_server_gpu/__init__.py
index 44733b154096255c3ce06e1be29d50d3e662269a..cc736a52e82637509d7d2a49efe9685f47c99e16 100644
--- a/python/paddle_serving_server_gpu/__init__.py
+++ b/python/paddle_serving_server_gpu/__init__.py
@@ -68,6 +68,10 @@ def serve_args():
         type=int,
         default=512 * 1024 * 1024,
         help="Limit sizes of messages")
+    parser.add_argument(
+        "--use_multilang",
+        action='store_true',
+        help="Use Multi-language-service")
     return parser.parse_args()
 
 
diff --git a/python/paddle_serving_server_gpu/serve.py b/python/paddle_serving_server_gpu/serve.py
index 309896a876bda5fc9b1baceb089242baa6d77dc5..0769039ef9955aea69af7ab84142a1f735ce4697 100644
--- a/python/paddle_serving_server_gpu/serve.py
+++ b/python/paddle_serving_server_gpu/serve.py
@@ -54,17 +54,26 @@ def start_gpu_card_model(index, gpuid, args):  # pylint: disable=doc-string-miss
     op_seq_maker.add_op(general_infer_op)
     op_seq_maker.add_op(general_response_op)
 
-    server = serving.Server()
-    server.set_op_sequence(op_seq_maker.get_op_sequence())
-    server.set_num_threads(thread_num)
-    server.set_memory_optimize(mem_optim)
-    server.set_ir_optimize(ir_optim)
-    server.set_max_body_size(max_body_size)
-
-    server.load_model_config(model)
-    server.prepare_server(workdir=workdir, port=port, device=device)
-    if gpuid >= 0:
-        server.set_gpuid(gpuid)
+    use_multilang = args.use_multilang
+    if use_multilang:
+        server = serving.MultiLangServer()
+        server.set_op_sequence(op_seq_maker.get_op_sequence())
+        server.load_model_config(model)
+        server.prepare_server(workdir=workdir, port=port, device=device)
+        if gpuid >= 0:
+            raise ValueError("gpuid can not >= 0 in MultiLangServer")
+    else:
+        server = serving.Server()
+        server.set_op_sequence(op_seq_maker.get_op_sequence())
+        server.set_num_threads(thread_num)
+        server.set_memory_optimize(mem_optim)
+        server.set_ir_optimize(ir_optim)
+        server.set_max_body_size(max_body_size)
+
+        server.load_model_config(model)
+        server.prepare_server(workdir=workdir, port=port, device=device)
+        if gpuid >= 0:
+            server.set_gpuid(gpuid)
     server.run_server()