diff --git a/comfy/model_management.py b/comfy/model_management.py
index ff963eb8e..74ac07de0 100644
--- a/comfy/model_management.py
+++ b/comfy/model_management.py
@@ -1983,6 +1983,14 @@ def supports_nvfp4_compute(device=None):
     if props.major < 10:
         return False
 
+    # cuBLAS FP4 matmul kernels require CUDA 13+, see #11864
+    try:
+        cuda_version_major = int(torch.version.cuda.split(".")[0])
+    except (AttributeError, ValueError):
+        return False
+    if cuda_version_major < 13:
+        return False
+
     return True
 
 def supports_mxfp8_compute(device=None):