From 9548a1887fe14e553c5db2c2a76e59fa79fd3ef4 Mon Sep 17 00:00:00 2001
From: Marceli Fylcek <marceli.fylcek@intel.com>
Date: Sun, 14 Jun 2026 10:14:35 +0300
Subject: [PATCH] [XPU] Support int4 group_size=32 W4A16 MoE (#45136)

Signed-off-by: Marceli Fylcek <marceli.fylcek@intel.com>
Co-authored-by: Kunshang Ji <kunshang.ji@intel.com>
---
 vllm/model_executor/layers/fused_moe/experts/xpu_moe.py | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/vllm/model_executor/layers/fused_moe/experts/xpu_moe.py b/vllm/model_executor/layers/fused_moe/experts/xpu_moe.py
index 00829a0f70..fe86e2b35f 100644
--- a/vllm/model_executor/layers/fused_moe/experts/xpu_moe.py
+++ b/vllm/model_executor/layers/fused_moe/experts/xpu_moe.py
@@ -19,6 +19,7 @@ from vllm.model_executor.layers.quantization.utils.quant_utils import (
     kFp8Static128BlockSym,
     kFp8StaticTensorSym,
     kInt4Static,
+    kInt4Static32,
     kMxfp4Static,
     kMxfp8Dynamic,
     kMxfp8Static,
@@ -302,7 +303,10 @@ class XPUExpertsWNA16(XPUExperts):
         weight_key: QuantKey | None,
         activation_key: QuantKey | None,
     ) -> bool:
-        return (weight_key, activation_key) == (kInt4Static, None)
+        return (weight_key, activation_key) in (
+            (kInt4Static, None),
+            (kInt4Static32, None),
+        )
 
 
 class XPUExpertsMxFp4(XPUExperts):