diff --git a/vllm/envs.py b/vllm/envs.py
index 03a22e4b2c7e..12f4664862a4 100755
--- a/vllm/envs.py
+++ b/vllm/envs.py
@@ -32,7 +32,7 @@
     VLLM_CONFIG_ROOT: str = os.path.expanduser("~/.config/vllm")
     VLLM_USAGE_STATS_SERVER: str = "https://stats.vllm.ai"
     VLLM_NO_USAGE_STATS: bool = False
-    VLLM_DISABLE_FLASHINFER_PREFILL: bool = False
+    VLLM_DISABLE_FLASHINFER_PREFILL: bool = True
     VLLM_DO_NOT_TRACK: bool = False
     VLLM_USAGE_SOURCE: str = ""
     VLLM_CONFIGURE_LOGGING: int = 1
@@ -557,7 +557,7 @@ def get_vllm_port() -> Optional[int]:
     "VLLM_NO_USAGE_STATS":
     lambda: os.environ.get("VLLM_NO_USAGE_STATS", "0") == "1",
     "VLLM_DISABLE_FLASHINFER_PREFILL":
-    lambda: os.environ.get("VLLM_DISABLE_FLASHINFER_PREFILL", "0") == "1",
+    lambda: os.environ.get("VLLM_DISABLE_FLASHINFER_PREFILL", "1") == "1",
     "VLLM_DO_NOT_TRACK":
     lambda: (os.environ.get("VLLM_DO_NOT_TRACK", None) or os.environ.get(
         "DO_NOT_TRACK", None) or "0") == "1",