diff --git a/requirements-dev.txt b/requirements-dev.txt
index ad60f499..eb775e24 100644
--- a/requirements-dev.txt
+++ b/requirements-dev.txt
@@ -2,7 +2,7 @@
 -r requirements.txt
 modelscope
 openai
-pytest >= 6.0
+pytest >= 6.0,<9.0.0
 pytest-asyncio
 pytest-mock
 lm-eval[api] @ git+https://github.com/EleutherAI/lm-evaluation-harness.git@206b7722158f58c35b7ffcd53b035fdbdda5126d 
diff --git a/tests/e2e/multicard/test_qwen3_next.py b/tests/e2e/multicard/test_qwen3_next.py
index e19ab1e1..ee213a8f 100644
--- a/tests/e2e/multicard/test_qwen3_next.py
+++ b/tests/e2e/multicard/test_qwen3_next.py
@@ -20,6 +20,8 @@
 
 Run `pytest tests/e2e/multicard/test_qwen3_next.py`.
 """
+import pytest
+
 from tests.e2e.conftest import VllmRunner
 
 
@@ -57,6 +59,7 @@ def test_models_distributed_Qwen3_NEXT_TP4_FULL_DECODE_ONLY():
         del vllm_model
 
 
+@pytest.mark.skip(reason="TODO: fix the test case later.")
 def test_models_distributed_Qwen3_NEXT_MTP_TP4_SIMILARITY():
     example_prompts = [
         "Hello, my name is",