diff --git a/python/sglang/test/kit_matched_stop.py b/python/sglang/test/kits/matched_stop_kit.py
similarity index 100%
rename from python/sglang/test/kit_matched_stop.py
rename to python/sglang/test/kits/matched_stop_kit.py
diff --git a/python/sglang/test/kits/radix_cache_server_kit.py b/python/sglang/test/kits/radix_cache_server_kit.py
new file mode 100644
index 000000000..12caa2b8c
--- /dev/null
+++ b/python/sglang/test/kits/radix_cache_server_kit.py
@@ -0,0 +1,50 @@
+import random
+
+import requests
+
+
+def gen_radix_tree(num_nodes=400, chunk_len=256):
+    num0 = num_nodes // 2
+    num1 = num_nodes - num0
+    nodes = [{"input_ids": [37] * 117, "decode_len": 217}]
+    for _ in range(num0):
+        parent = random.choice(nodes)
+        unique_len = random.randint(0, chunk_len)
+        decode_len = random.randint(0, chunk_len)
+        token_id = random.randint(0, 32000)
+        child = {
+            "input_ids": parent["input_ids"] + [token_id] * unique_len,
+            "decode_len": decode_len,
+        }
+        nodes.append(child)
+
+    while num1 > 0:
+        num_branch = random.randint(1, min(num1, 10))
+        parent = random.choice(nodes)
+        for _ in range(num_branch):
+            unique_len = random.randint(0, chunk_len)
+            decode_len = random.randint(0, chunk_len)
+            token_id = random.randint(0, 32000)
+            child = {
+                "input_ids": parent["input_ids"] + [token_id] * unique_len,
+                "decode_len": decode_len,
+            }
+            nodes.append(child)
+
+        num1 -= num_branch
+
+    random.shuffle(nodes)
+    return nodes
+
+
+def run_radix_attention_test(base_url: str):
+    nodes = gen_radix_tree()
+    data = {
+        "input_ids": [node["input_ids"] for node in nodes],
+        "sampling_params": [
+            {"max_new_tokens": node["decode_len"], "temperature": 0} for node in nodes
+        ],
+    }
+
+    res = requests.post(base_url + "/generate", json=data)
+    assert res.status_code == 200
diff --git a/test/srt/openai_server/validation/test_matched_stop.py b/test/srt/openai_server/validation/test_matched_stop.py
index 12496dbb6..62c586bac 100644
--- a/test/srt/openai_server/validation/test_matched_stop.py
+++ b/test/srt/openai_server/validation/test_matched_stop.py
@@ -2,7 +2,7 @@ import unittest
 
 from sglang.srt.sampling.sampling_params import MAX_LEN, get_max_seq_length
 from sglang.srt.utils import kill_process_tree
-from sglang.test.kit_matched_stop import MatchedStopMixin
+from sglang.test.kits.matched_stop_kit import MatchedStopMixin
 from sglang.test.test_utils import (
     DEFAULT_MODEL_NAME_FOR_TEST,
     DEFAULT_URL_FOR_TEST,
diff --git a/test/srt/run_suite.py b/test/srt/run_suite.py
index 19e9ed3f3..825c9f863 100644
--- a/test/srt/run_suite.py
+++ b/test/srt/run_suite.py
@@ -78,7 +78,7 @@ suites = {
         TestFile("test_deterministic.py", 320),
         TestFile("test_eagle_infer_a.py", 370),
         TestFile("test_eagle_infer_b.py", 700),
-        TestFile("test_eagle_infer_beta.py", 300),
+        TestFile("test_eagle_infer_beta.py", 90),
         TestFile("test_ebnf_constrained.py", 108),
         TestFile("test_eval_fp8_accuracy.py", 303),
         TestFile("test_fa3.py", 376),
diff --git a/test/srt/test_eagle_infer_beta.py b/test/srt/test_eagle_infer_beta.py
index 6e3d34471..f1c3cd258 100644
--- a/test/srt/test_eagle_infer_beta.py
+++ b/test/srt/test_eagle_infer_beta.py
@@ -4,7 +4,8 @@ from types import SimpleNamespace
 from sglang.srt.environ import envs
 from sglang.srt.utils import kill_process_tree
 from sglang.test.few_shot_gsm8k import run_eval
-from sglang.test.kit_matched_stop import MatchedStopMixin
+from sglang.test.kits.matched_stop_kit import MatchedStopMixin
+from sglang.test.kits.radix_cache_server_kit import run_radix_attention_test
 from sglang.test.test_utils import (
     DEFAULT_EAGLE_DRAFT_MODEL_FOR_TEST,
     DEFAULT_EAGLE_TARGET_MODEL_FOR_TEST,
@@ -65,6 +66,9 @@ class TestEagleServerBase(CustomTestCase, MatchedStopMixin):
     def tearDownClass(cls):
         kill_process_tree(cls.process.pid)
 
+    def test_radix_attention(self):
+        run_radix_attention_test(self.base_url)
+
     def test_gsm8k(self):
         args = SimpleNamespace(
             num_shots=5,
diff --git a/test/srt/test_radix_attention.py b/test/srt/test_radix_attention.py
index 66f948621..c8c3f539e 100644
--- a/test/srt/test_radix_attention.py
+++ b/test/srt/test_radix_attention.py
@@ -1,9 +1,7 @@
-import os
-import random
 import unittest
 
-import requests
-
+from sglang.srt.environ import envs
+from sglang.test.kits.radix_cache_server_kit import run_radix_attention_test
 from sglang.test.test_utils import (
     DEFAULT_SMALL_MODEL_NAME_FOR_TEST,
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
@@ -15,52 +13,6 @@ from sglang.test.test_utils import (
 )
 
 
-def gen_radix_tree(num_nodes=400, chunk_len=256):
-    num0 = num_nodes // 2
-    num1 = num_nodes - num0
-    nodes = [{"input_ids": [37] * 117, "decode_len": 217}]
-    for _ in range(num0):
-        parent = random.choice(nodes)
-        unique_len = random.randint(0, chunk_len)
-        decode_len = random.randint(0, chunk_len)
-        token_id = random.randint(0, 32000)
-        child = {
-            "input_ids": parent["input_ids"] + [token_id] * unique_len,
-            "decode_len": decode_len,
-        }
-        nodes.append(child)
-
-    while num1 > 0:
-        num_branch = random.randint(1, min(num1, 10))
-        parent = random.choice(nodes)
-        for _ in range(num_branch):
-            unique_len = random.randint(0, chunk_len)
-            decode_len = random.randint(0, chunk_len)
-            token_id = random.randint(0, 32000)
-            child = {
-                "input_ids": parent["input_ids"] + [token_id] * unique_len,
-                "decode_len": decode_len,
-            }
-            nodes.append(child)
-
-        num1 -= num_branch
-
-    random.shuffle(nodes)
-    return nodes
-
-
-def run_test(base_url, nodes):
-    data = {
-        "input_ids": [node["input_ids"] for node in nodes],
-        "sampling_params": [
-            {"max_new_tokens": node["decode_len"], "temperature": 0} for node in nodes
-        ],
-    }
-
-    res = requests.post(base_url + "/generate", json=data)
-    assert res.status_code == 200
-
-
 class TestRadixCacheFCFS(CustomTestCase):
     @classmethod
     def setUpClass(cls):
@@ -85,8 +37,7 @@ class TestRadixCacheFCFS(CustomTestCase):
         kill_process_tree(cls.process.pid)
 
     def test_radix_attention(self):
-        nodes = gen_radix_tree()
-        run_test(self.base_url, nodes)
+        run_radix_attention_test(self.base_url)
 
 
 @unittest.skipIf(is_in_ci(), "To reduce the CI execution time.")
@@ -132,5 +83,5 @@ class TestRadixCacheNonOverlapLPM(TestRadixCacheFCFS):
 
 
 if __name__ == "__main__":
-    os.environ["SGLANG_TEST_RETRACT"] = "true"
+    envs.SGLANG_TEST_RETRACT.set(True)
     unittest.main()