diff --git a/apps/shark_studio/api/llm.py b/apps/shark_studio/api/llm.py
index 1a03b817ff..2c08a2a6e1 100644
--- a/apps/shark_studio/api/llm.py
+++ b/apps/shark_studio/api/llm.py
@@ -106,6 +106,7 @@ def compile(self) -> None:
             frontend="torch",
             external_weight_file=self.external_weight_file,
             write_to=self.vmfb_name,
+            extra_args=["--iree-global-opt-enable-quantized-matmul-reassociation"]
         )
         # TODO: delete the temp file