diff --git a/python/sglang/srt/layers/quantization/modelopt_quant.py b/python/sglang/srt/layers/quantization/modelopt_quant.py
index efa58209b..330fa6c21 100755
--- a/python/sglang/srt/layers/quantization/modelopt_quant.py
+++ b/python/sglang/srt/layers/quantization/modelopt_quant.py
@@ -2027,6 +2027,17 @@ class ModelOptNvFp4FusedMoEMethod(FusedMoEMethodBase):
         else:
             w13_input_scale = layer.w13_input_scale.max(dim=-1).values.to(torch.float32)
             w2_input_scale = layer.w2_input_scale
+            # w13_input_scale/w2_input_scale are loaded GLOBAL (num_experts) because
+            # create_weights tags them _sglang_require_global_experts, while
+            # w13_weight_scale_2 is EP-local (num_local_experts). Under EP>1 this
+            # non-flashinfer path must slice the global input scales down to this
+            # rank's local experts so the gemm1/gemm2 alpha products line up
+            # (mirrors the cutedsl branch above). No-op for moe_ep_size == 1.
+            if layer.moe_ep_size > 1 and w13_input_scale.shape[0] == layer.num_experts:
+                start = layer.moe_ep_rank * layer.num_local_experts
+                end = start + layer.num_local_experts
+                w13_input_scale = w13_input_scale[start:end]
+                w2_input_scale = w2_input_scale[start:end]
 
         if self.quant_config.use_per_token_activation:
             # FlashInfer computes activation scales dynamically per token, so