diff --git a/volc_limit_sync.py b/volc_limit_sync.py
index f9dbebc..1285d9d 100644
--- a/volc_limit_sync.py
+++ b/volc_limit_sync.py
@@ -648,11 +648,15 @@ def _check_request_block(model_name):
             return None
 
         for entry in llm_router.model_list:
-            if entry.get("model_name") != model_name:
-                continue
-            # 已被限额踢出的条目跳过（路由层已匹配不到）
-            if "_daily_limit_backup" in entry:
-                continue
+            entry_name = entry.get("model_name") or ""
+            # 匹配请求名；被限额软踢的条目（原名被重命名为 {name}-limit-reached）
+            # 也参与判定，保证超限模型无论是否已踢都返回 429，而非 router 400
+            if entry_name != model_name:
+                if not (
+                    "_daily_limit_backup" in entry
+                    and entry_name == f"{model_name}-limit-reached"
+                ):
+                    continue
 
             params = entry.get("litellm_params", {})
             raw_model = params.get("model", "")