diff --git a/litellm/llms/azure/responses/transformation.py b/litellm/llms/azure/responses/transformation.py index 0dd5e87e4ba..cfd0e96639f 100644 --- a/litellm/llms/azure/responses/transformation.py +++ b/litellm/llms/azure/responses/transformation.py @@ -6,6 +6,7 @@ from openai.types.responses import ResponseReasoningItem from litellm._logging import verbose_logger from litellm.litellm_core_utils.url_utils import encode_url_path_segment +from litellm.llms.azure.chat.gpt_5_transformation import AzureOpenAIGPT5Config from litellm.llms.azure.common_utils import BaseAzureLLM from litellm.llms.openai.responses.transformation import OpenAIResponsesAPIConfig from litellm.types.llms.openai import * @@ -29,6 +30,14 @@ class AzureOpenAIResponsesAPIConfig(OpenAIResponsesAPIConfig): def custom_llm_provider(self) -> LlmProviders: return LlmProviders.AZURE + @staticmethod + def _supports_reasoning_effort_none(model: str) -> bool: + return AzureOpenAIGPT5Config._supports_reasoning_effort_level(model, "none") + + @staticmethod + def _effort_resolves_to_none(model: str, effort: str | None) -> bool: + return AzureOpenAIGPT5Config.effort_resolves_to_none(model, effort) + def get_supported_openai_params(self, model: str) -> list: """ Azure Responses API does not support context_management (compaction). diff --git a/litellm/model_prices_and_context_window_backup.json b/litellm/model_prices_and_context_window_backup.json index 7f5038e3073..2459ed940e0 100644 --- a/litellm/model_prices_and_context_window_backup.json +++ b/litellm/model_prices_and_context_window_backup.json @@ -7157,6 +7157,53 @@ "supports_xhigh_reasoning_effort": true, "supports_minimal_reasoning_effort": false }, + "azure/gpt-6-astra": { + "cache_creation_input_token_cost": 1.25e-05, + "cache_creation_input_token_cost_above_272k_tokens": 2.5e-05, + "cache_read_input_token_cost": 1e-06, + "cache_read_input_token_cost_above_272k_tokens": 2e-06, + "input_cost_per_token": 1e-05, + "input_cost_per_token_above_272k_tokens": 2e-05, + "litellm_provider": "azure", + "max_input_tokens": 922000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 5e-05, + "output_cost_per_token_above_272k_tokens": 7.5e-05, + "search_context_cost_per_query": { + "search_context_size_high": 0.01, + "search_context_size_low": 0.01, + "search_context_size_medium": 0.01 + }, + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/responses" + ], + "supported_modalities": [ + "text", + "image" + ], + "supported_output_modalities": [ + "text" + ], + "supports_computer_use": true, + "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_minimal_reasoning_effort": false, + "supports_native_streaming": true, + "supports_none_reasoning_effort": true, + "supports_parallel_function_calling": true, + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_vision": true, + "supports_web_search": true, + "supports_xhigh_reasoning_effort": true + }, "azure/us/gpt-5.6": { "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_272k_tokens": 1.375e-05, @@ -7376,6 +7423,53 @@ "supports_xhigh_reasoning_effort": true, "supports_minimal_reasoning_effort": false }, + "azure/us/gpt-6-astra": { + "cache_creation_input_token_cost": 1.375e-05, + "cache_creation_input_token_cost_above_272k_tokens": 2.75e-05, + "cache_read_input_token_cost": 1.1e-06, + "cache_read_input_token_cost_above_272k_tokens": 2.2e-06, + "input_cost_per_token": 1.1e-05, + "input_cost_per_token_above_272k_tokens": 2.2e-05, + "litellm_provider": "azure", + "max_input_tokens": 922000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 5.5e-05, + "output_cost_per_token_above_272k_tokens": 8.25e-05, + "search_context_cost_per_query": { + "search_context_size_high": 0.01, + "search_context_size_low": 0.01, + "search_context_size_medium": 0.01 + }, + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/responses" + ], + "supported_modalities": [ + "text", + "image" + ], + "supported_output_modalities": [ + "text" + ], + "supports_computer_use": true, + "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_minimal_reasoning_effort": false, + "supports_native_streaming": true, + "supports_none_reasoning_effort": true, + "supports_parallel_function_calling": true, + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_vision": true, + "supports_web_search": true, + "supports_xhigh_reasoning_effort": true + }, "azure/eu/gpt-5.6": { "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_272k_tokens": 1.375e-05, diff --git a/model_prices_and_context_window.json b/model_prices_and_context_window.json index 7f5038e3073..2459ed940e0 100644 --- a/model_prices_and_context_window.json +++ b/model_prices_and_context_window.json @@ -7157,6 +7157,53 @@ "supports_xhigh_reasoning_effort": true, "supports_minimal_reasoning_effort": false }, + "azure/gpt-6-astra": { + "cache_creation_input_token_cost": 1.25e-05, + "cache_creation_input_token_cost_above_272k_tokens": 2.5e-05, + "cache_read_input_token_cost": 1e-06, + "cache_read_input_token_cost_above_272k_tokens": 2e-06, + "input_cost_per_token": 1e-05, + "input_cost_per_token_above_272k_tokens": 2e-05, + "litellm_provider": "azure", + "max_input_tokens": 922000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 5e-05, + "output_cost_per_token_above_272k_tokens": 7.5e-05, + "search_context_cost_per_query": { + "search_context_size_high": 0.01, + "search_context_size_low": 0.01, + "search_context_size_medium": 0.01 + }, + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/responses" + ], + "supported_modalities": [ + "text", + "image" + ], + "supported_output_modalities": [ + "text" + ], + "supports_computer_use": true, + "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_minimal_reasoning_effort": false, + "supports_native_streaming": true, + "supports_none_reasoning_effort": true, + "supports_parallel_function_calling": true, + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_vision": true, + "supports_web_search": true, + "supports_xhigh_reasoning_effort": true + }, "azure/us/gpt-5.6": { "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_272k_tokens": 1.375e-05, @@ -7376,6 +7423,53 @@ "supports_xhigh_reasoning_effort": true, "supports_minimal_reasoning_effort": false }, + "azure/us/gpt-6-astra": { + "cache_creation_input_token_cost": 1.375e-05, + "cache_creation_input_token_cost_above_272k_tokens": 2.75e-05, + "cache_read_input_token_cost": 1.1e-06, + "cache_read_input_token_cost_above_272k_tokens": 2.2e-06, + "input_cost_per_token": 1.1e-05, + "input_cost_per_token_above_272k_tokens": 2.2e-05, + "litellm_provider": "azure", + "max_input_tokens": 922000, + "max_output_tokens": 128000, + "max_tokens": 128000, + "mode": "chat", + "output_cost_per_token": 5.5e-05, + "output_cost_per_token_above_272k_tokens": 8.25e-05, + "search_context_cost_per_query": { + "search_context_size_high": 0.01, + "search_context_size_low": 0.01, + "search_context_size_medium": 0.01 + }, + "supported_endpoints": [ + "/v1/chat/completions", + "/v1/responses" + ], + "supported_modalities": [ + "text", + "image" + ], + "supported_output_modalities": [ + "text" + ], + "supports_computer_use": true, + "supports_function_calling": true, + "supports_max_reasoning_effort": true, + "supports_minimal_reasoning_effort": false, + "supports_native_streaming": true, + "supports_none_reasoning_effort": true, + "supports_parallel_function_calling": true, + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_system_messages": true, + "supports_tool_choice": true, + "supports_vision": true, + "supports_web_search": true, + "supports_xhigh_reasoning_effort": true + }, "azure/eu/gpt-5.6": { "cache_creation_input_token_cost": 6.875e-06, "cache_creation_input_token_cost_above_272k_tokens": 1.375e-05, diff --git a/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py b/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py index 0f8084643ea..df680b7cb0e 100644 --- a/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py +++ b/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py @@ -2008,6 +2008,49 @@ def test_generic_cost_per_token_azure_gpt56(_local_model_cost_map, assert round(completion_cost, 10) == round(output_cost * completion_tokens, 10) +@pytest.mark.parametrize("model,zone_multiplier", [("azure/gpt-6-astra", 1.0), ("azure/us/gpt-6-astra", 1.1)]) +@pytest.mark.parametrize( + "prompt_tokens,input_side_multiplier,output_multiplier", + [(100000, 1.0, 1.0), (300000, 2.0, 1.5)], +) +def test_generic_cost_per_token_azure_gpt_6_astra_foundry_price_sheet( + _local_model_cost_map, + model, + zone_multiplier, + prompt_tokens, + input_side_multiplier, + output_multiplier, +): + """Microsoft Foundry sells gpt-6-astra at the OpenAI rates: $10 input, $1 cache read, $12.50 cache write, + $50 output per 1M tokens on Standard Global, with the input side doubling and output 1.5x above 272K + prompt tokens. Standard US Data Zone carries the usual 10% uplift on every rate. + """ + cached_tokens = 50000 + cache_write_tokens = 40000 + text_tokens = prompt_tokens - cached_tokens - cache_write_tokens + completion_tokens = 1000 + usage = Usage( + prompt_tokens=prompt_tokens, + completion_tokens=completion_tokens, + total_tokens=prompt_tokens + completion_tokens, + prompt_tokens_details=PromptTokensDetailsWrapper( + cached_tokens=cached_tokens, cache_write_tokens=cache_write_tokens + ), + ) + + prompt_cost, completion_cost = generic_cost_per_token( + model=model, + usage=usage, + custom_llm_provider="azure", + ) + + input_side = zone_multiplier * input_side_multiplier + assert prompt_cost == pytest.approx( + input_side * (text_tokens * 1e-5 + cached_tokens * 1e-6 + cache_write_tokens * 1.25e-5) + ) + assert completion_cost == pytest.approx(zone_multiplier * output_multiplier * completion_tokens * 5e-5) + + @pytest.mark.parametrize( "model,expected_none,expected_xhigh,expected_minimal", [ diff --git a/tests/test_litellm/llms/azure/chat/test_azure_gpt5_transformation.py b/tests/test_litellm/llms/azure/chat/test_azure_gpt5_transformation.py index bd0f16a695b..3b3ef1a9cd4 100644 --- a/tests/test_litellm/llms/azure/chat/test_azure_gpt5_transformation.py +++ b/tests/test_litellm/llms/azure/chat/test_azure_gpt5_transformation.py @@ -348,3 +348,31 @@ def test_azure_gpt_6_astra_takes_the_reasoning_series_request_shape(): assert params["max_completion_tokens"] == 100 assert "max_tokens" not in params assert params["reasoning_effort"] == "max" + + +@pytest.mark.parametrize("model", ["azure/gpt-6-astra", "azure/us/gpt-6-astra"]) +def test_azure_gpt6_astra_reasoning_effort_none_unlocks_temperature(config: AzureOpenAIGPT5Config, model: str): + """Foundry's gpt-6-astra accepts reasoning_effort='none' and, only then, a non-default + temperature (verified live against a Foundry deployment), unlike OpenAI's gpt-6-astra.""" + params = config.map_openai_params( + non_default_params={"temperature": 0.2, "reasoning_effort": "none"}, + optional_params={}, + model=model, + drop_params=False, + api_version="2025-04-01-preview", + ) + assert params["temperature"] == 0.2 + assert params["reasoning_effort"] == "none" + + +@pytest.mark.parametrize("model", ["azure/gpt-6-astra", "azure/us/gpt-6-astra"]) +def test_azure_gpt6_astra_rejects_reasoning_effort_minimal(config: AzureOpenAIGPT5Config, model: str): + """Foundry's gpt-6-astra lists none, low, medium, high, xhigh and max but not minimal.""" + with pytest.raises(litellm.utils.UnsupportedParamsError): + config.map_openai_params( + non_default_params={"reasoning_effort": "minimal"}, + optional_params={}, + model=model, + drop_params=False, + api_version="2025-04-01-preview", + ) diff --git a/tests/test_litellm/llms/azure/response/test_azure_transformation.py b/tests/test_litellm/llms/azure/response/test_azure_transformation.py index f6bbf685f26..0cac2705ab0 100644 --- a/tests/test_litellm/llms/azure/response/test_azure_transformation.py +++ b/tests/test_litellm/llms/azure/response/test_azure_transformation.py @@ -1,11 +1,10 @@ from copy import deepcopy -from unittest.mock import patch +from unittest.mock import MagicMock, patch import pytest - -from unittest.mock import MagicMock - +import litellm +from litellm.litellm_core_utils.get_model_cost_map import get_model_cost_map from litellm.llms.azure.responses.o_series_transformation import ( AzureOpenAIOSeriesResponsesAPIConfig, ) @@ -613,3 +612,39 @@ class TestAzureResponsesAPIConfig: assert result["tools"][0] is tool assert "anyOf" in result["tools"][0]["parameters"] + + +@pytest.fixture() +def local_model_cost_map(monkeypatch: pytest.MonkeyPatch) -> None: + """Pin the bundled cost map: the published map lags a key added in this repo.""" + monkeypatch.setenv("LITELLM_LOCAL_MODEL_COST_MAP", "True") + monkeypatch.setattr(litellm, "model_cost", get_model_cost_map(url=litellm.model_cost_map_url)) + litellm.add_known_models(model_cost_map=litellm.model_cost) + + +def test_azure_responses_gpt6_astra_reasoning_effort_none_unlocks_temperature(local_model_cost_map: None): + """Foundry's gpt-6-astra accepts reasoning.effort='none' with a non-default temperature + while OpenAI's gpt-6-astra does not, so the gate must read the azure/ cost-map entry + for the bare deployment name rather than OpenAI's.""" + params = AzureOpenAIResponsesAPIConfig().map_openai_params( + response_api_optional_params=ResponsesAPIOptionalRequestParams( + temperature=0.2, + reasoning={"effort": "none"}, + ), + model="gpt-6-astra", + drop_params=False, + ) + assert params["temperature"] == 0.2 + assert params["reasoning"] == {"effort": "none"} + + +def test_azure_responses_gpt6_astra_rejects_temperature_while_reasoning(local_model_cost_map: None): + with pytest.raises(litellm.UnsupportedParamsError): + AzureOpenAIResponsesAPIConfig().map_openai_params( + response_api_optional_params=ResponsesAPIOptionalRequestParams( + temperature=0.2, + reasoning={"effort": "low"}, + ), + model="gpt-6-astra", + drop_params=False, + ) diff --git a/tests/test_litellm/router_utils/test_reasoning_effort_capability.py b/tests/test_litellm/router_utils/test_reasoning_effort_capability.py index 7b2e45ab3ed..f181370455d 100644 --- a/tests/test_litellm/router_utils/test_reasoning_effort_capability.py +++ b/tests/test_litellm/router_utils/test_reasoning_effort_capability.py @@ -388,3 +388,21 @@ class TestGpt6AstraAdvertisesItsDocumentedLevels: "xhigh", "max", ) + + @pytest.mark.parametrize("model", ["azure/gpt-6-astra", "azure/us/gpt-6-astra"]) + def test_a_foundry_deployment_also_advertises_none(self, local_model_cost_map, model): + """Microsoft Foundry serves the same model but its API accepts reasoning_effort none + (verified live: 200 with zero reasoning tokens, and it unlocks temperature), which + OpenAI's rejects, so an Azure deployment offers none on top of low through max.""" + from litellm.utils import _get_model_info_helper + + model_info = dict(_get_model_info_helper(model=model, custom_llm_provider="azure")) + + assert resolve_supported_reasoning_efforts(model_info, deployment_is_mapped=True) == ( + "none", + "low", + "medium", + "high", + "xhigh", + "max", + )