diff --git a/tests/attention/test_xqa_batch_decode.py b/tests/attention/test_xqa_batch_decode.py
index 27fff7846ea..97f50710bf5 100644
--- a/tests/attention/test_xqa_batch_decode.py
+++ b/tests/attention/test_xqa_batch_decode.py
@@ -581,7 +581,7 @@ def test_xqa_batch_decode(
 
 @pytest.mark.skipif(
     get_compute_capability(torch.device(device="cuda"))[0] not in [12],
-    reason="XQA with NVFP4 KV is only supported on SM120 GPUs",
+    reason="XQA with NVFP4 KV is only supported on SM120/SM121 GPUs",
 )
 @pytest.mark.parametrize(
     "batch_size,q_len_per_req,page_size,num_kv_heads,head_grp_size",