diff --git a/dion/dion2.py b/dion/dion2.py
index 8837aea..3ad2c93 100644
--- a/dion/dion2.py
+++ b/dion/dion2.py
@@ -61,7 +61,7 @@ def __init__(
         flatten: bool = False,
         use_triton: bool = False,
         use_polar_express: bool = True,
-        use_gram_newton_schulz: bool = True,
+        use_gram_newton_schulz: bool = False,
         newton_schulz_func: Optional[Callable] = None,
         verbose: bool = False,
     ):
diff --git a/dion/megabatch_base.py b/dion/megabatch_base.py
index 519e6ce..47e8062 100644
--- a/dion/megabatch_base.py
+++ b/dion/megabatch_base.py
@@ -34,7 +34,7 @@ def __init__(
         distributed_mesh: Optional[Union[DeviceMesh, ProcessGroup]],
         algo_name: str,
         defaults: dict,
-        use_gram_newton_schulz: bool = True,
+        use_gram_newton_schulz: bool = False,
         use_triton: bool = False,
         use_polar_express: bool = True,
         newton_schulz_func: Optional[Callable] = None,
diff --git a/dion/muon.py b/dion/muon.py
index a320052..4a96b9d 100644
--- a/dion/muon.py
+++ b/dion/muon.py
@@ -61,7 +61,7 @@ def __init__(
         nesterov: bool = False,
         adjust_lr: Optional[str] = "spectral_norm",
         flatten: bool = False,
-        use_gram_newton_schulz: bool = True,
+        use_gram_newton_schulz: bool = False,
         use_triton: bool = False,
         use_polar_express: bool = True,
         newton_schulz_func: Optional[Callable] = None,
diff --git a/dion/normuon.py b/dion/normuon.py
index a419186..34ccce3 100644
--- a/dion/normuon.py
+++ b/dion/normuon.py
@@ -64,7 +64,7 @@ def __init__(
         nesterov: bool = False,
         adjust_lr: Optional[str] = "spectral_norm",
         flatten: bool = False,
-        use_gram_newton_schulz: bool = True,
+        use_gram_newton_schulz: bool = False,
         use_triton: bool = False,
         use_polar_express: bool = True,
         newton_schulz_func: Optional[Callable] = None,