diff --git a/python/sglang/bench_offline_throughput.py b/python/sglang/benchmark/offline_throughput.py
similarity index 100%
rename from python/sglang/bench_offline_throughput.py
rename to python/sglang/benchmark/offline_throughput.py
diff --git a/python/sglang/test/test_utils.py b/python/sglang/test/test_utils.py
index d2f898364..91406f717 100644
--- a/python/sglang/test/test_utils.py
+++ b/python/sglang/test/test_utils.py
@@ -1616,7 +1616,7 @@ def run_bench_offline_throughput(model, other_args):
     command = [
         "python3",
         "-m",
-        "sglang.bench_offline_throughput",
+        "sglang.benchmark.offline_throughput",
         "--num-prompts",
         "1",
         "--dataset-name",
diff --git a/test/manual/perf/test_bench_one_batch_1gpu.py b/test/manual/perf/test_bench_one_batch_1gpu.py
index 5a0c39383..b70818f21 100644
--- a/test/manual/perf/test_bench_one_batch_1gpu.py
+++ b/test/manual/perf/test_bench_one_batch_1gpu.py
@@ -31,7 +31,7 @@ class TestBenchOneBatch1GPU(CustomTestCase):
         command = [
             "python3",
             "-m",
-            "sglang.bench_offline_throughput",
+            "sglang.benchmark.offline_throughput",
             "--num-prompts",
             "1",
             "--dataset-name",
diff --git a/test/registered/core/test_srt_engine.py b/test/registered/core/test_srt_engine.py
index f5c6ecfc8..c3e86e5d6 100644
--- a/test/registered/core/test_srt_engine.py
+++ b/test/registered/core/test_srt_engine.py
@@ -10,7 +10,7 @@ import unittest
 import torch
 
 import sglang as sgl
-from sglang.bench_offline_throughput import BenchArgs, throughput_test
+from sglang.benchmark.offline_throughput import BenchArgs, throughput_test
 from sglang.srt.server_args import ServerArgs
 from sglang.srt.utils.hf_transformers_utils import get_tokenizer
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci