Skip to content

The agent engineering platform.

revert(openai): switch model from nano to mini when using flex (#34394)

Mason Daughertyc286c06f164c
diff --git a/libs/partners/openai/tests/integration_tests/chat_models/test_base.py b/libs/partners/openai/tests/integration_tests/chat_models/test_base.py
index 2588c33fd5..02475b5a8b 100644
--- a/libs/partners/openai/tests/integration_tests/chat_models/test_base.py
+++ b/libs/partners/openai/tests/integration_tests/chat_models/test_base.py
@@ -239,7 +239,7 @@ async def test_openai_abatch_tags(use_responses_api: bool) -> None:
def test_openai_invoke() -> None:
"""Test invoke tokens from ChatOpenAI."""
llm = ChatOpenAI(
- model="gpt-5-mini",
+ model="gpt-5-nano",
service_tier="flex", # Also test service_tier
max_retries=3, # Add retries for 503 capacity errors
)
@@ -272,7 +272,7 @@ def test_openai_invoke() -> None:
def test_stream() -> None:
"""Test streaming tokens from OpenAI."""
llm = ChatOpenAI(
- model="gpt-5-mini",
+ model="gpt-5-nano",
service_tier="flex", # Also test service_tier
max_retries=3, # Add retries for 503 capacity errors
)
@@ -399,7 +399,7 @@ async def test_astream() -> None:
@pytest.mark.parametrize("streaming", [False, True])
def test_flex_usage_responses(streaming: bool) -> None:
llm = ChatOpenAI(
- model="gpt-5-mini",
+ model="gpt-5-nano",
service_tier="flex",
max_retries=3,
use_responses_api=True,