From 86483ca7749b3d7a2ae16283a7896c203983f1ef Mon Sep 17 00:00:00 2001
From: tomeras91 <57313761+tomeras91@users.noreply.github.com>
Date: Thu, 5 Mar 2026 19:49:05 +0200
Subject: [PATCH] [Bugfix] Disable FlashInfer TRTLLM BF16 path for non-gated
 MoE (#36146)

Signed-off-by: Tomer Asida <57313761+tomeras91@users.noreply.github.com>
---
 .../layers/fused_moe/flashinfer_trtllm_moe.py               | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/vllm/model_executor/layers/fused_moe/flashinfer_trtllm_moe.py b/vllm/model_executor/layers/fused_moe/flashinfer_trtllm_moe.py
index 6765e3613..d04e040c8 100644
--- a/vllm/model_executor/layers/fused_moe/flashinfer_trtllm_moe.py
+++ b/vllm/model_executor/layers/fused_moe/flashinfer_trtllm_moe.py
@@ -25,12 +25,12 @@ def _supports_current_device() -> bool:
 
 
 def _supports_no_act_and_mul() -> bool:
-    """Supports non-gated MoE."""
-    return True
+    """BF16 kernels do not support non-gated MoE"""
+    return False
 
 
 def _supports_activation(activation: MoEActivation) -> bool:
-    return activation in [MoEActivation.SILU, MoEActivation.RELU2_NO_MUL]
+    return activation in [MoEActivation.SILU]
 
 
 def _supports_routing_method_bf16(