diff --git a/python/sglang/srt/server_args.py b/python/sglang/srt/server_args.py
index bc17db6f89e0..e8fdda3f3622 100644
--- a/python/sglang/srt/server_args.py
+++ b/python/sglang/srt/server_args.py
@@ -1815,6 +1815,21 @@ def _handle_speculative_decoding(self):
         ):
             self.speculative_draft_model_revision = "main"
 
+        # Avoid using flashinfer_trtllm for speculative MoE runner backend by default
+        # TODO: Remove this block after verifying no accuracy regression with flashinfer_trtllm speculative backend
+        from sglang.srt.layers.moe.utils import MoeRunnerBackend
+
+        if self.speculative_moe_runner_backend is None:
+            self.speculative_moe_runner_backend = (
+                "auto"
+                if self.moe_runner_backend == "flashinfer_trtllm"
+                else self.moe_runner_backend
+            )
+        else:
+            assert not MoeRunnerBackend(
+                self.speculative_moe_runner_backend
+            ).is_flashinfer_trtllm(), "Currently speculative MoE runner backend cannot be flashinfer_trtllm for risk in some draft models."
+
         if self.speculative_algorithm == "NEXTN":
             self.speculative_algorithm = "EAGLE"
 
diff --git a/test/srt/test_deepseek_v3_fp4_4gpu.py b/test/srt/test_deepseek_v3_fp4_4gpu.py
index 1f236f0953c8..2b4f94889cd3 100644
--- a/test/srt/test_deepseek_v3_fp4_4gpu.py
+++ b/test/srt/test_deepseek_v3_fp4_4gpu.py
@@ -172,7 +172,7 @@ def test_bs_1_speed(self):
                 f"{speed=:.2f} token/s\n"
             )
 
-        self.assertGreater(acc_length, 2.04)
+        self.assertGreater(acc_length, 2.65)
         self.assertGreater(speed, 150)