diff --git a/torchao/prototype/mx_formats/README.md b/torchao/prototype/mx_formats/README.md
index b5ba91326d..5323727a8b 100644
--- a/torchao/prototype/mx_formats/README.md
+++ b/torchao/prototype/mx_formats/README.md
@@ -92,6 +92,7 @@ config = MXLinearConfig(
     scale_calculation_mode=scale_calculation_mode,
 )
 quantize_(m, config)
+m = torch.compile(m, fullgraph=True)
 
 # training loop (not shown)
 ```