diff --git a/python/sglang/srt/layers/moe/fused_moe_triton/layer.py b/python/sglang/srt/layers/moe/fused_moe_triton/layer.py
index 5be46dec0..1c2aa6f81 100644
--- a/python/sglang/srt/layers/moe/fused_moe_triton/layer.py
+++ b/python/sglang/srt/layers/moe/fused_moe_triton/layer.py
@@ -373,7 +373,12 @@ class FusedMoE(torch.nn.Module):
             self.moe_runner_config.inplace = False
 
         self.should_fuse_routed_scaling_factor_in_topk = (
-            isinstance(self.quant_method, ModelOptNvFp4FusedMoEMethod)
+            (
+                isinstance(self.quant_method, ModelOptNvFp4FusedMoEMethod)
+                and not getattr(
+                    self.quant_method, "_moe_runner_backend", get_moe_runner_backend()
+                ).is_marlin()
+            )
             or (
                 isinstance(self.quant_method, Fp8MoEMethod)
                 and (