diff --git a/tests/lora/test_tokenizer_group.py b/tests/lora/test_tokenizer_group.py
index daa39b2a3dba1..d225a3f7d6c06 100644
--- a/tests/lora/test_tokenizer_group.py
+++ b/tests/lora/test_tokenizer_group.py
@@ -17,6 +17,7 @@ async def test_tokenizer_group_lora(sql_lora_files, tokenizer_group_type):
         tokenizer_id="gpt2",
         enable_lora=True,
         max_num_seqs=1,
+        max_loras=1,
         max_input_length=None,
     )
     lora_request = LoRARequest("1", 1, sql_lora_files)
@@ -53,3 +54,22 @@ def test_get_lora_tokenizer(sql_lora_files, tmp_path):
     lora_request = LoRARequest("1", 1, str(tmp_path))
     tokenizer = get_lora_tokenizer(lora_request)
     assert not tokenizer
+
+
+@pytest.mark.parametrize("enable_lora", [True, False])
+@pytest.mark.parametrize("max_num_seqs", [1, 2])
+@pytest.mark.parametrize("max_loras", [1, 2])
+def test_lora_tokenizers(enable_lora, max_num_seqs, max_loras):
+    tokenizer_group = get_tokenizer_group(
+        get_tokenizer_pool_config(None),
+        tokenizer_id="gpt2",
+        enable_lora=enable_lora,
+        max_num_seqs=max_num_seqs,
+        max_loras=max_loras,
+        max_input_length=None,
+    )
+    if enable_lora:
+        assert tokenizer_group.lora_tokenizers.capacity == max(
+            max_num_seqs, max_loras)
+    else:
+        assert tokenizer_group.lora_tokenizers.capacity == 0