diff --git a/vllm/engine/arg_utils.py b/vllm/engine/arg_utils.py
index 27f62b0008578..31aa8c5908719 100644
--- a/vllm/engine/arg_utils.py
+++ b/vllm/engine/arg_utils.py
@@ -626,8 +626,8 @@ def add_cli_args(parser: FlexibleArgumentParser) -> FlexibleArgumentParser:
             type=int,
             default=EngineArgs.max_cpu_loras,
             help=('Maximum number of LoRAs to store in CPU memory. '
-                  'Must be >= than max_num_seqs. '
-                  'Defaults to max_num_seqs.'))
+                  'Must be >= than max_loras. '
+                  'Defaults to max_loras.'))
         parser.add_argument(
             '--fully-sharded-loras',
             action='store_true',