diff --git a/tests/unit/llms/anthropic/messages/adapters/test_anthropic_messages_adapters_transformation.py b/tests/unit/llms/anthropic/messages/adapters/test_anthropic_messages_adapters_transformation.py index bad7b3f1750..a76d5e82a37 100644 --- a/tests/unit/llms/anthropic/messages/adapters/test_anthropic_messages_adapters_transformation.py +++ b/tests/unit/llms/anthropic/messages/adapters/test_anthropic_messages_adapters_transformation.py @@ -759,7 +759,7 @@ def test_translate_openai_reasoning_items_are_not_duplicated_when_thinking_block def _replayed_reasoning_request(tools: list[dict[str, object]]) -> AnthropicMessagesRequest: return AnthropicMessagesRequest( - model="azure/gpt-5.4-nano", + model="azure/gpt-5.6", max_tokens=64, messages=[ AnthropicMessagesUserMessageParam(role="user", content="Which city is colder?"), @@ -5820,7 +5820,7 @@ _RETRY_REPLY: Final = { "object": "response", "created_at": 1700000000, "status": "completed", - "model": "gpt-5.4-nano", + "model": "gpt-5.6", "output": [ { "type": "message", @@ -5868,7 +5868,7 @@ async def test_async_messages_retry_once_without_reasoning_the_backend_cannot_de with caplog.at_level(logging.WARNING, logger="LiteLLM"): response: Final = await litellm.anthropic.messages.acreate( - model="azure/gpt-5.4-nano", + model="azure/gpt-5.6", api_base=_RETRY_API_BASE, api_key="fake-key", api_version="2025-04-01-preview", @@ -5890,7 +5890,7 @@ async def test_async_messages_retry_once_without_reasoning_the_backend_cannot_de assert retry_warnings == ( "Backend rejected replayed encrypted reasoning, retrying once without it. " "Enable the encrypted_content_affinity pre-call check to drop such blocks before the call. " - "provider=azure model=azure/gpt-5.4-nano status=400 code=invalid_encrypted_content", + "provider=azure model=azure/gpt-5.6 status=400 code=invalid_encrypted_content", ) assert all( "gAAA_foreign" not in message and "litellm_encrypted_reasoning:" not in message for message in warnings @@ -5906,7 +5906,7 @@ def test_sync_messages_retry_once_without_reasoning_the_backend_cannot_decrypt( ) response: Final = litellm.anthropic.messages.create( - model="azure/gpt-5.4-nano", + model="azure/gpt-5.6", api_base=_RETRY_API_BASE, api_key="fake-key", api_version="2025-04-01-preview", @@ -5930,7 +5930,7 @@ async def test_async_messages_with_num_retries_send_the_rejected_reasoning_only_ ) await litellm.anthropic.messages.acreate( - model="azure/gpt-5.4-nano", + model="azure/gpt-5.6", api_base=_RETRY_API_BASE, api_key="fake-key", api_version="2025-04-01-preview", @@ -5953,7 +5953,7 @@ def test_sync_messages_with_global_num_retries_send_the_rejected_reasoning_once_ ) litellm.anthropic.messages.create( - model="azure/gpt-5.4-nano", + model="azure/gpt-5.6", api_base=_RETRY_API_BASE, api_key="fake-key", api_version="2025-04-01-preview", @@ -5981,7 +5981,7 @@ async def test_async_messages_retry_a_server_error_with_the_reasoning_and_drop_i ) response: Final = await litellm.anthropic.messages.acreate( - model="azure/gpt-5.4-nano", + model="azure/gpt-5.6", api_base=_RETRY_API_BASE, api_key="fake-key", api_version="2025-04-01-preview", @@ -6004,7 +6004,7 @@ def test_sync_messages_retry_a_server_error_with_the_reasoning_and_drop_it_only_ ) response: Final = litellm.anthropic.messages.create( - model="azure/gpt-5.4-nano", + model="azure/gpt-5.6", api_base=_RETRY_API_BASE, api_key="fake-key", api_version="2025-04-01-preview", @@ -6063,7 +6063,7 @@ async def test_async_messages_log_only_the_outcome_of_the_turn_sent_again( with pytest.raises(litellm.BadRequestError) if logged_rejections else nullcontext(): await litellm.anthropic.messages.acreate( - model="azure/gpt-5.4-nano", + model="azure/gpt-5.6", api_base=_RETRY_API_BASE, api_key="fake-key", api_version="2025-04-01-preview", @@ -6089,7 +6089,7 @@ def test_sync_messages_log_only_the_outcome_of_the_turn_sent_again( with pytest.raises(litellm.BadRequestError) if logged_rejections else nullcontext(): litellm.anthropic.messages.create( - model="azure/gpt-5.4-nano", + model="azure/gpt-5.6", api_base=_RETRY_API_BASE, api_key="fake-key", api_version="2025-04-01-preview", @@ -6124,7 +6124,7 @@ async def test_messages_do_not_retry_other_rejections( with pytest.raises(litellm.BadRequestError): await litellm.anthropic.messages.acreate( - model="azure/gpt-5.4-nano", + model="azure/gpt-5.6", api_base=_RETRY_API_BASE, api_key="fake-key", api_version="2025-04-01-preview", @@ -6277,7 +6277,7 @@ async def test_streamed_reasoning_summaries_stay_with_their_own_encrypted_item( ) stream: Final = await litellm.anthropic.messages.acreate( - model="azure/gpt-5.4-nano", + model="azure/gpt-5.6", api_base=_RETRY_API_BASE, api_key="fake-key", api_version="2025-04-01-preview", diff --git a/tests/unit/proxy/guardrails/test_deferred_guardrail_logging.py b/tests/unit/proxy/guardrails/test_deferred_guardrail_logging.py index e9e0a585949..e21dc857005 100644 --- a/tests/unit/proxy/guardrails/test_deferred_guardrail_logging.py +++ b/tests/unit/proxy/guardrails/test_deferred_guardrail_logging.py @@ -330,7 +330,7 @@ async def test_deferred_slot_keeps_the_innermost_wrapper_result(): async def test_deferred_anthropic_messages_bridged_to_the_responses_api_logs_the_provider_usage( respx_mock: respx.MockRouter, monkeypatch: pytest.MonkeyPatch ): - """/v1/messages on an Azure gpt-5.4+ deployment with function tools runs three nested + """/v1/messages on an Azure deployment with function tools and explicit reasoning runs three nested wrappers: anthropic_messages, the chat adapter's acompletion, and the Responses bridge acompletion hands the call to, which retags the call as ``responses``. With logging deferred for a post-call guardrail the stored closure must carry the innermost provider @@ -382,6 +382,7 @@ async def test_deferred_anthropic_messages_bridged_to_the_responses_api_logs_the response: Final = await litellm.anthropic_messages( model="azure/gpt-5.4-nano", + reasoning_effort="medium", messages=[{"role": "user", "content": "hi"}], max_tokens=16, tools=[