diff --git a/vllm/model_executor/layers/fused_moe/fused_moe.py b/vllm/model_executor/layers/fused_moe/fused_moe.py
index c412f695ae76..1985e8612da3 100644
--- a/vllm/model_executor/layers/fused_moe/fused_moe.py
+++ b/vllm/model_executor/layers/fused_moe/fused_moe.py
@@ -1127,6 +1127,7 @@ def flashinfer_fused_moe_blockscale_fp8(
         tile_tokens_dim=_get_tile_tokens_dim(x.shape[0], top_k,
                                              global_num_experts),
         routing_method_type=2,  # DeepSeek-styled routing method
+        use_shuffled_weight=False,
     )