diff --git a/libs/partners/openai/tests/unit_tests/chat_models/test_azure.py b/libs/partners/openai/tests/unit_tests/chat_models/test_azure.py
index 41cba5445e3..492e54a67c2 100644
--- a/libs/partners/openai/tests/unit_tests/chat_models/test_azure.py
+++ b/libs/partners/openai/tests/unit_tests/chat_models/test_azure.py
@@ -85,6 +85,65 @@ def test_structured_output_old_model() -> None:
     assert "response_format" not in llm.steps[0].kwargs  # type: ignore
 
 
+def test_azure_client_caching() -> None:
+    """Test that the Azure OpenAI client is cached."""
+    llm1 = AzureChatOpenAI(
+        azure_deployment="gpt-35-turbo",
+        api_version="2023-05-15",
+        azure_endpoint="https://test.openai.azure.com/",
+    )
+    llm2 = AzureChatOpenAI(
+        azure_deployment="gpt-35-turbo",
+        api_version="2023-05-15",
+        azure_endpoint="https://test.openai.azure.com/",
+    )
+    assert llm1.root_client._client is llm2.root_client._client
+
+    # Different endpoint should create a different client
+    llm3 = AzureChatOpenAI(
+        azure_deployment="gpt-35-turbo",
+        api_version="2023-05-15",
+        azure_endpoint="https://different.openai.azure.com/",
+    )
+    assert llm1.root_client._client is not llm3.root_client._client
+
+    # Same endpoint with timeout=None should reuse the client
+    llm4 = AzureChatOpenAI(
+        azure_deployment="gpt-35-turbo",
+        api_version="2023-05-15",
+        azure_endpoint="https://test.openai.azure.com/",
+        timeout=None,
+    )
+    assert llm1.root_client._client is llm4.root_client._client
+
+    # Different timeout should create a different client
+    llm5 = AzureChatOpenAI(
+        azure_deployment="gpt-35-turbo",
+        api_version="2023-05-15",
+        azure_endpoint="https://test.openai.azure.com/",
+        timeout=3,
+    )
+    assert llm1.root_client._client is not llm5.root_client._client
+
+    # httpx.Timeout object should create a different client
+    llm6 = AzureChatOpenAI(
+        azure_deployment="gpt-35-turbo",
+        api_version="2023-05-15",
+        azure_endpoint="https://test.openai.azure.com/",
+        timeout=httpx.Timeout(timeout=60.0, connect=5.0),
+    )
+    assert llm1.root_client._client is not llm6.root_client._client
+
+    # Tuple timeout should create a different client
+    llm7 = AzureChatOpenAI(
+        azure_deployment="gpt-35-turbo",
+        api_version="2023-05-15",
+        azure_endpoint="https://test.openai.azure.com/",
+        timeout=(5, 1),
+    )
+    assert llm1.root_client._client is not llm7.root_client._client
+
+
 def test_max_completion_tokens_in_payload() -> None:
     llm = AzureChatOpenAI(
         azure_deployment="o1-mini",
@@ -101,3 +160,4 @@ def test_max_completion_tokens_in_payload() -> None:
         "max_completion_tokens": 300,
     }
 
+