diff --git a/runtimes/huggingface/mlserver_huggingface/runtime.py b/runtimes/huggingface/mlserver_huggingface/runtime.py
index de1c25bf8..90ed9f398 100644
--- a/runtimes/huggingface/mlserver_huggingface/runtime.py
+++ b/runtimes/huggingface/mlserver_huggingface/runtime.py
@@ -19,6 +19,7 @@
 from mlserver_huggingface.codecs import MultiStringRequestCodec
 from transformers.pipelines import SUPPORTED_TASKS
 from optimum.pipelines import SUPPORTED_TASKS as SUPPORTED_OPTIMUM_TASKS
+from mlserver.logging import logger
 
 
 class HuggingFaceRuntime(MLModel):
@@ -57,6 +58,12 @@ def __init__(self, settings: ModelSettings):
                     ),
                 )
 
+        if settings.max_batch_size != self.hf_settings.batch_size:
+            logger.warning(
+                f"hf batch_size: {self.hf_settings.batch_size} is different "
+                f"from MLServer max_batch_size: {settings.max_batch_size}"
+            )
+
         super().__init__(settings)
 
     async def load(self) -> bool: