diff --git a/python/sglang/srt/model_executor/model_runner.py b/python/sglang/srt/model_executor/model_runner.py
index 8b06d2cea..134effc5a 100644
--- a/python/sglang/srt/model_executor/model_runner.py
+++ b/python/sglang/srt/model_executor/model_runner.py
@@ -233,7 +233,10 @@ class ModelRunner:
 
         # Prepare the vllm model config
         monkey_patch_vllm_dummy_weight_loader()
-        self.load_config = LoadConfig(load_format=self.server_args.load_format)
+        self.load_config = LoadConfig(
+            load_format=self.server_args.load_format,
+            download_dir=self.server_args.download_dir,
+        )
         self.vllm_model_config = VllmModelConfig(
             model=self.server_args.model_path,
             quantization=self.server_args.quantization,
diff --git a/python/sglang/srt/server_args.py b/python/sglang/srt/server_args.py
index 7003d2c53..c20aac847 100644
--- a/python/sglang/srt/server_args.py
+++ b/python/sglang/srt/server_args.py
@@ -64,6 +64,7 @@ class ServerArgs:
     random_seed: Optional[int] = None
     constrained_json_whitespace_pattern: Optional[str] = None
     watchdog_timeout: float = 300
+    download_dir: Optional[str] = None
 
     # Logging
     log_level: str = "info"
@@ -405,6 +406,12 @@ class ServerArgs:
             default=ServerArgs.watchdog_timeout,
             help="Set watchdog timeout in seconds. If a forward batch takes longer than this, the server will crash to prevent hanging.",
         )
+        parser.add_argument(
+            "--download-dir",
+            type=str,
+            default=ServerArgs.download_dir,
+            help="Model download directory.",
+        )
 
         # Logging
         parser.add_argument(