diff --git a/python/sglang/srt/managers/controller/tp_worker.py b/python/sglang/srt/managers/controller/tp_worker.py
index 9b223263b6..30e209b228 100644
--- a/python/sglang/srt/managers/controller/tp_worker.py
+++ b/python/sglang/srt/managers/controller/tp_worker.py
@@ -101,8 +101,7 @@ def __init__(
         self.int_token_logit_bias = torch.tensor(
             get_int_token_logit_bias(self.tokenizer, self.model_config.vocab_size)
         )
-        if server_args.random_seed is not None:
-            set_random_seed(server_args.random_seed)
+        set_random_seed(server_args.random_seed)
 
         # Print info
         logger.info(
diff --git a/python/sglang/srt/server_args.py b/python/sglang/srt/server_args.py
index b733a3f782..35cc024112 100644
--- a/python/sglang/srt/server_args.py
+++ b/python/sglang/srt/server_args.py
@@ -32,7 +32,7 @@ class ServerArgs:
     # Other runtime options
     tp_size: int = 1
     stream_interval: int = 8
-    random_seed: Optional[int] = None
+    random_seed: int = 42
 
     # Logging
     log_level: str = "info"
@@ -194,7 +194,7 @@ def add_cli_args(parser: argparse.ArgumentParser):
             "--random-seed",
             type=int,
             default=ServerArgs.random_seed,
-            help="The global random seed of the server.",
+            help="Random seed.",
         )
         parser.add_argument(
             "--log-level",