test(openai): mark tests flaky (#33750)

see:
https://github.com/langchain-ai/langchain/actions/runs/18921929210/job/54020065079#step:10:560

libs/partners/openai/tests/integration_tests/chat_models/test_base.py

@@ -1003,6 +1003,7 @@ def test_audio_input_modality() -> None:
     assert "audio" in output.additional_kwargs
 
 
+@pytest.mark.flaky(retries=3, delay=1)
 def test_prediction_tokens() -> None:
     code = dedent(
         """

libs/partners/openai/tests/integration_tests/chat_models/test_base_standard.py

@@ -88,6 +88,7 @@ class TestOpenAIStandard(ChatModelIntegrationTests):
         # For now, we test with filename in OpenAI-specific tests
         return False
 
+    @pytest.mark.flaky(retries=3, delay=1)
     def test_openai_pdf_inputs(self, model: BaseChatModel) -> None:
         """Test that the model can process PDF inputs."""
         url = "https://www.w3.org/WAI/ER/tests/xhtml/testfiles/resources/pdf/dummy.pdf"

libs/partners/openai/tests/integration_tests/chat_models/test_responses_standard.py

@@ -54,6 +54,7 @@ class TestOpenAIResponses(TestOpenAIStandard):
         input_ = "What was the 3rd highest building in 2000?"
         return _invoke(llm, input_, stream)
 
+    @pytest.mark.flaky(retries=3, delay=1)
     def test_openai_pdf_inputs(self, model: BaseChatModel) -> None:
         """Test that the model can process PDF inputs."""
         super().test_openai_pdf_inputs(model)