feat(cost-map): add azure_ai/gpt-6-astra Foundry pricing

A gpt-6-astra deployment on a Foundry project reached through the
azure_ai route had no cost map entry of its own, so it resolved to the
OpenAI gpt-6-astra card: missing from the azure_ai/* wildcard listing,
flex and priority prices and /v1/batch it does not sell, and no none
reasoning effort. Add azure_ai/gpt-6-astra mirroring the
azure/gpt-6-astra Standard Global sheet the way azure_ai/gpt-5.5 mirrors
azure/gpt-5.5, and extend the cost, reasoning-effort, and wildcard
listing tests to the Foundry route.

litellm/model_prices_and_context_window_backup.json

@@ -3485,6 +3485,49 @@
         "supports_response_schema": true,
         "supports_tool_choice": true
     },
+    "azure_ai/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_ai",
+        "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,
+        "source": "https://ai.azure.com/catalog/models/gpt-6-astra",
+        "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_ai/gpt-5.5": {
         "deprecation_date": "2027-10-26",
         "cache_read_input_token_cost": 5e-07,

model_prices_and_context_window.json

@@ -3485,6 +3485,49 @@
         "supports_response_schema": true,
         "supports_tool_choice": true
     },
+    "azure_ai/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_ai",
+        "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,
+        "source": "https://ai.azure.com/catalog/models/gpt-6-astra",
+        "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_ai/gpt-5.5": {
         "deprecation_date": "2027-10-26",
         "cache_read_input_token_cost": 5e-07,

tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py

@@ -2008,7 +2008,14 @@ 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(
+    "model,custom_llm_provider,zone_multiplier",
+    [
+        ("azure/gpt-6-astra", "azure", 1.0),
+        ("azure/us/gpt-6-astra", "azure", 1.1),
+        ("azure_ai/gpt-6-astra", "azure_ai", 1.0),
+    ],
+)
 @pytest.mark.parametrize(
     "prompt_tokens,input_side_multiplier,output_multiplier",
     [(100000, 1.0, 1.0), (300000, 2.0, 1.5)],
@@ -2016,6 +2023,7 @@ def test_generic_cost_per_token_azure_gpt56(_local_model_cost_map,
 def test_generic_cost_per_token_azure_gpt_6_astra_foundry_price_sheet(
     _local_model_cost_map,
     model,
+    custom_llm_provider,
     zone_multiplier,
     prompt_tokens,
     input_side_multiplier,
@@ -2023,7 +2031,8 @@ def test_generic_cost_per_token_azure_gpt_6_astra_foundry_price_sheet(
 ):
     """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.
+    prompt tokens. Standard US Data Zone carries the usual 10% uplift on every rate. A Foundry
+    deployment reached through the azure_ai route bills the same Standard Global sheet.
     """
     cached_tokens = 50000
     cache_write_tokens = 40000
@@ -2041,7 +2050,7 @@ def test_generic_cost_per_token_azure_gpt_6_astra_foundry_price_sheet(
     prompt_cost, completion_cost = generic_cost_per_token(
         model=model,
         usage=usage,
-        custom_llm_provider="azure",
+        custom_llm_provider=custom_llm_provider,
     )
 
     input_side = zone_multiplier * input_side_multiplier

tests/test_litellm/proxy/auth/test_model_checks.py

@@ -857,6 +857,25 @@ def test_add_known_models_refreshes_models_by_provider_for_wildcard_expansion():
         litellm.add_known_models(model_cost_map={})
     assert fake_model not in litellm.models_by_provider["vertex_ai"]
 
+
+def test_azure_ai_wildcard_lists_the_foundry_gpt_6_astra_entry(monkeypatch):
+    """A Foundry (azure_ai) deployment of gpt-6-astra only shows up under an azure_ai/* wildcard
+    when the cost map carries its own azure_ai/ entry; the azure/ entry from the OpenAI-on-Azure
+    price sheet never reaches the Foundry provider list (LIT-7081)."""
+    import litellm
+    from litellm.proxy.auth.model_checks import get_known_models_from_wildcard
+
+    monkeypatch.setenv("LITELLM_LOCAL_MODEL_COST_MAP", "True")
+    foundry_key = "azure_ai/gpt-6-astra"
+    local_entry = litellm.get_model_cost_map(url="")[foundry_key]
+    try:
+        litellm.add_known_models(model_cost_map={foundry_key: local_entry})
+        assert foundry_key in get_known_models_from_wildcard("azure_ai/*")
+    finally:
+        litellm.azure_ai_models.discard(foundry_key)
+        litellm.add_known_models(model_cost_map={})
+
+
 def test_get_complete_model_list_drops_no_default_models_sentinel():
     from litellm.proxy.auth.model_checks import get_complete_model_list
 

tests/test_litellm/router_utils/test_reasoning_effort_capability.py

@@ -389,14 +389,22 @@ class TestGpt6AstraAdvertisesItsDocumentedLevels:
             "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):
+    @pytest.mark.parametrize(
+        "model,custom_llm_provider",
+        [
+            ("azure/gpt-6-astra", "azure"),
+            ("azure/us/gpt-6-astra", "azure"),
+            ("azure_ai/gpt-6-astra", "azure_ai"),
+        ],
+    )
+    def test_a_foundry_deployment_also_advertises_none(self, local_model_cost_map, model, custom_llm_provider):
         """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."""
+        OpenAI's rejects, so an Azure deployment offers none on top of low through max, whether
+        it is reached through the azure route or the azure_ai (Foundry) route."""
         from litellm.utils import _get_model_info_helper
 
-        model_info = dict(_get_model_info_helper(model=model, custom_llm_provider="azure"))
+        model_info = dict(_get_model_info_helper(model=model, custom_llm_provider=custom_llm_provider))
 
         assert resolve_supported_reasoning_efforts(model_info, deployment_is_mapped=True) == (
             "none",