diff --git a/vllm/model_executor/layers/fused_moe/layer.py b/vllm/model_executor/layers/fused_moe/layer.py
index 815e5c5fb7..15806ca4f8 100644
--- a/vllm/model_executor/layers/fused_moe/layer.py
+++ b/vllm/model_executor/layers/fused_moe/layer.py
@@ -349,7 +349,9 @@ def FusedMoE(
         topk_group=topk_group,
         custom_routing_function=custom_routing_function,
         scoring_func=scoring_func,
-        routed_scaling_factor=routed_scaling_factor,
+        routed_scaling_factor=routed_scaling_factor
+        if not apply_routed_scale_to_output
+        else 1.0,
         swiglu_limit=swiglu_limit,
         # TODO get from router? needs to be truncated?
         e_score_correction_bias=e_score_correction_bias,