diff --git a/python/sglang/srt/distributed/device_communicators/custom_all_reduce.py b/python/sglang/srt/distributed/device_communicators/custom_all_reduce.py
index ac308df63..b14211e17 100644
--- a/python/sglang/srt/distributed/device_communicators/custom_all_reduce.py
+++ b/python/sglang/srt/distributed/device_communicators/custom_all_reduce.py
@@ -342,7 +342,8 @@ def dispatch_custom_allreduce():
     On AMD with 1-stage AR enabled, use sglang's CustomAllreduce.
     Otherwise use AiterCustomAllreduce if available.
 
-    Set SGLANG_OPT_USE_CUSTOM_ALL_REDUCE_V2=1 to use the JIT-compiled v2 implementation.
+    On CUDA, the JIT-compiled v2 implementation is used by default.
+    Set SGLANG_OPT_USE_CUSTOM_ALL_REDUCE_V2=0 to fall back to the legacy CustomAllreduce.
     """
     if _is_cuda and envs.SGLANG_OPT_USE_CUSTOM_ALL_REDUCE_V2.get():
         from .custom_all_reduce_v2 import CustomAllReduceV2
diff --git a/python/sglang/srt/environ.py b/python/sglang/srt/environ.py
index 3167b8ee2..c87321c35 100644
--- a/python/sglang/srt/environ.py
+++ b/python/sglang/srt/environ.py
@@ -448,7 +448,7 @@ class Envs:
     # Set to 1: force enable (even without --enable-deterministic-inference)
     # Set to 0: force disable (use default Aiter AR even with --enable-deterministic-inference)
     SGLANG_USE_1STAGE_ALLREDUCE = EnvBool(False)
-    SGLANG_OPT_USE_CUSTOM_ALL_REDUCE_V2 = EnvBool(False)
+    SGLANG_OPT_USE_CUSTOM_ALL_REDUCE_V2 = EnvBool(True)
     SGLANG_FLASHINFER_PREFILL_SPLIT_TILE_SIZE = EnvInt(4096)
     SGLANG_FLASHINFER_DECODE_SPLIT_TILE_SIZE = EnvInt(2048)
     SGLANG_TRITON_PREFILL_TRUNCATION_ALIGN_SIZE = EnvInt(4096)