fix(model-map): advertise max reasoning for GPT-5.6 and Azure Astra

This commit is contained in:
Tin 2026-10-23 16:52:33 -04:00
parent f0f15404d5
commit e98f96e769
8 changed files with 110 additions and 52 deletions

View file

@ -4124,7 +4124,7 @@
],
"supports_computer_use": true,
"supports_function_calling": true,
"supports_max_reasoning_effort": false,
"supports_max_reasoning_effort": true,
"supports_minimal_reasoning_effort": false,
"supports_native_streaming": true,
"supports_none_reasoning_effort": false,
@ -8351,6 +8351,7 @@
"supports_pdf_input": true,
"supports_prompt_caching": true,
"supports_reasoning": true,
"supports_max_reasoning_effort": true,
"supports_response_schema": true,
"supports_system_messages": true,
"supports_tool_choice": true,
@ -8415,6 +8416,7 @@
"supports_pdf_input": true,
"supports_prompt_caching": true,
"supports_reasoning": true,
"supports_max_reasoning_effort": true,
"supports_response_schema": true,
"supports_system_messages": true,
"supports_tool_choice": true,
@ -8479,6 +8481,7 @@
"supports_pdf_input": true,
"supports_prompt_caching": true,
"supports_reasoning": true,
"supports_max_reasoning_effort": true,
"supports_response_schema": true,
"supports_system_messages": true,
"supports_tool_choice": true,
@ -8543,6 +8546,7 @@
"supports_pdf_input": true,
"supports_prompt_caching": true,
"supports_reasoning": true,
"supports_max_reasoning_effort": true,
"supports_response_schema": true,
"supports_system_messages": true,
"supports_tool_choice": true,
@ -8607,6 +8611,7 @@
"supports_pdf_input": true,
"supports_prompt_caching": true,
"supports_reasoning": true,
"supports_max_reasoning_effort": true,
"supports_response_schema": true,
"supports_system_messages": true,
"supports_tool_choice": true,
@ -8670,6 +8675,7 @@
"supports_pdf_input": true,
"supports_prompt_caching": true,
"supports_reasoning": true,
"supports_max_reasoning_effort": true,
"supports_response_schema": true,
"supports_system_messages": true,
"supports_tool_choice": true,
@ -8714,7 +8720,7 @@
],
"supports_computer_use": true,
"supports_function_calling": true,
"supports_max_reasoning_effort": false,
"supports_max_reasoning_effort": true,
"supports_minimal_reasoning_effort": false,
"supports_native_streaming": true,
"supports_none_reasoning_effort": false,
@ -8763,7 +8769,7 @@
],
"supports_computer_use": true,
"supports_function_calling": true,
"supports_max_reasoning_effort": false,
"supports_max_reasoning_effort": true,
"supports_minimal_reasoning_effort": false,
"supports_native_streaming": true,
"supports_none_reasoning_effort": false,
@ -9402,6 +9408,7 @@
"supports_pdf_input": true,
"supports_prompt_caching": true,
"supports_reasoning": true,
"supports_max_reasoning_effort": true,
"supports_response_schema": true,
"supports_system_messages": true,
"supports_tool_choice": true,
@ -9458,6 +9465,7 @@
"supports_pdf_input": true,
"supports_prompt_caching": true,
"supports_reasoning": true,
"supports_max_reasoning_effort": true,
"supports_response_schema": true,
"supports_system_messages": true,
"supports_tool_choice": true,
@ -9514,6 +9522,7 @@
"supports_pdf_input": true,
"supports_prompt_caching": true,
"supports_reasoning": true,
"supports_max_reasoning_effort": true,
"supports_response_schema": true,
"supports_system_messages": true,
"supports_tool_choice": true,
@ -9557,7 +9566,7 @@
],
"supports_computer_use": true,
"supports_function_calling": true,
"supports_max_reasoning_effort": false,
"supports_max_reasoning_effort": true,
"supports_minimal_reasoning_effort": false,
"supports_native_streaming": true,
"supports_none_reasoning_effort": false,
@ -9812,6 +9821,7 @@
"supports_pdf_input": true,
"supports_prompt_caching": true,
"supports_reasoning": true,
"supports_max_reasoning_effort": true,
"supports_response_schema": true,
"supports_system_messages": true,
"supports_tool_choice": true,
@ -9868,6 +9878,7 @@
"supports_pdf_input": true,
"supports_prompt_caching": true,
"supports_reasoning": true,
"supports_max_reasoning_effort": true,
"supports_response_schema": true,
"supports_system_messages": true,
"supports_tool_choice": true,
@ -9924,6 +9935,7 @@
"supports_pdf_input": true,
"supports_prompt_caching": true,
"supports_reasoning": true,
"supports_max_reasoning_effort": true,
"supports_response_schema": true,
"supports_system_messages": true,
"supports_tool_choice": true,
@ -33823,6 +33835,7 @@
"supports_minimal_reasoning_effort": false,
"supports_none_reasoning_effort": true,
"supports_reasoning": true,
"supports_max_reasoning_effort": true,
"supports_xhigh_reasoning_effort": true
},
"chatgpt/gpt-5.6-sol": {
@ -33843,6 +33856,7 @@
"supports_minimal_reasoning_effort": false,
"supports_none_reasoning_effort": true,
"supports_reasoning": true,
"supports_max_reasoning_effort": true,
"supports_xhigh_reasoning_effort": true
},
"chatgpt/gpt-5.6-terra": {
@ -33863,6 +33877,7 @@
"supports_minimal_reasoning_effort": false,
"supports_none_reasoning_effort": true,
"supports_reasoning": true,
"supports_max_reasoning_effort": true,
"supports_xhigh_reasoning_effort": true
},
"chatgpt/gpt-5.4": {
@ -37087,6 +37102,7 @@
"supports_prompt_cache_breakpoint": true,
"supports_prompt_caching": true,
"supports_reasoning": true,
"supports_max_reasoning_effort": true,
"supports_response_schema": true,
"supports_system_messages": true,
"supports_tool_choice": true,
@ -37162,6 +37178,7 @@
"supports_prompt_cache_breakpoint": true,
"supports_prompt_caching": true,
"supports_reasoning": true,
"supports_max_reasoning_effort": true,
"supports_response_schema": true,
"supports_system_messages": true,
"supports_tool_choice": true,
@ -37237,6 +37254,7 @@
"supports_prompt_cache_breakpoint": true,
"supports_prompt_caching": true,
"supports_reasoning": true,
"supports_max_reasoning_effort": true,
"supports_response_schema": true,
"supports_system_messages": true,
"supports_tool_choice": true,
@ -70354,7 +70372,7 @@
],
"supports_computer_use": true,
"supports_function_calling": true,
"supports_max_reasoning_effort": false,
"supports_max_reasoning_effort": true,
"supports_minimal_reasoning_effort": false,
"supports_native_streaming": true,
"supports_none_reasoning_effort": false,

View file

@ -4124,7 +4124,7 @@
],
"supports_computer_use": true,
"supports_function_calling": true,
"supports_max_reasoning_effort": false,
"supports_max_reasoning_effort": true,
"supports_minimal_reasoning_effort": false,
"supports_native_streaming": true,
"supports_none_reasoning_effort": false,
@ -8351,6 +8351,7 @@
"supports_pdf_input": true,
"supports_prompt_caching": true,
"supports_reasoning": true,
"supports_max_reasoning_effort": true,
"supports_response_schema": true,
"supports_system_messages": true,
"supports_tool_choice": true,
@ -8415,6 +8416,7 @@
"supports_pdf_input": true,
"supports_prompt_caching": true,
"supports_reasoning": true,
"supports_max_reasoning_effort": true,
"supports_response_schema": true,
"supports_system_messages": true,
"supports_tool_choice": true,
@ -8479,6 +8481,7 @@
"supports_pdf_input": true,
"supports_prompt_caching": true,
"supports_reasoning": true,
"supports_max_reasoning_effort": true,
"supports_response_schema": true,
"supports_system_messages": true,
"supports_tool_choice": true,
@ -8543,6 +8546,7 @@
"supports_pdf_input": true,
"supports_prompt_caching": true,
"supports_reasoning": true,
"supports_max_reasoning_effort": true,
"supports_response_schema": true,
"supports_system_messages": true,
"supports_tool_choice": true,
@ -8607,6 +8611,7 @@
"supports_pdf_input": true,
"supports_prompt_caching": true,
"supports_reasoning": true,
"supports_max_reasoning_effort": true,
"supports_response_schema": true,
"supports_system_messages": true,
"supports_tool_choice": true,
@ -8670,6 +8675,7 @@
"supports_pdf_input": true,
"supports_prompt_caching": true,
"supports_reasoning": true,
"supports_max_reasoning_effort": true,
"supports_response_schema": true,
"supports_system_messages": true,
"supports_tool_choice": true,
@ -8714,7 +8720,7 @@
],
"supports_computer_use": true,
"supports_function_calling": true,
"supports_max_reasoning_effort": false,
"supports_max_reasoning_effort": true,
"supports_minimal_reasoning_effort": false,
"supports_native_streaming": true,
"supports_none_reasoning_effort": false,
@ -8763,7 +8769,7 @@
],
"supports_computer_use": true,
"supports_function_calling": true,
"supports_max_reasoning_effort": false,
"supports_max_reasoning_effort": true,
"supports_minimal_reasoning_effort": false,
"supports_native_streaming": true,
"supports_none_reasoning_effort": false,
@ -9402,6 +9408,7 @@
"supports_pdf_input": true,
"supports_prompt_caching": true,
"supports_reasoning": true,
"supports_max_reasoning_effort": true,
"supports_response_schema": true,
"supports_system_messages": true,
"supports_tool_choice": true,
@ -9458,6 +9465,7 @@
"supports_pdf_input": true,
"supports_prompt_caching": true,
"supports_reasoning": true,
"supports_max_reasoning_effort": true,
"supports_response_schema": true,
"supports_system_messages": true,
"supports_tool_choice": true,
@ -9514,6 +9522,7 @@
"supports_pdf_input": true,
"supports_prompt_caching": true,
"supports_reasoning": true,
"supports_max_reasoning_effort": true,
"supports_response_schema": true,
"supports_system_messages": true,
"supports_tool_choice": true,
@ -9557,7 +9566,7 @@
],
"supports_computer_use": true,
"supports_function_calling": true,
"supports_max_reasoning_effort": false,
"supports_max_reasoning_effort": true,
"supports_minimal_reasoning_effort": false,
"supports_native_streaming": true,
"supports_none_reasoning_effort": false,
@ -9812,6 +9821,7 @@
"supports_pdf_input": true,
"supports_prompt_caching": true,
"supports_reasoning": true,
"supports_max_reasoning_effort": true,
"supports_response_schema": true,
"supports_system_messages": true,
"supports_tool_choice": true,
@ -9868,6 +9878,7 @@
"supports_pdf_input": true,
"supports_prompt_caching": true,
"supports_reasoning": true,
"supports_max_reasoning_effort": true,
"supports_response_schema": true,
"supports_system_messages": true,
"supports_tool_choice": true,
@ -9924,6 +9935,7 @@
"supports_pdf_input": true,
"supports_prompt_caching": true,
"supports_reasoning": true,
"supports_max_reasoning_effort": true,
"supports_response_schema": true,
"supports_system_messages": true,
"supports_tool_choice": true,
@ -33823,6 +33835,7 @@
"supports_minimal_reasoning_effort": false,
"supports_none_reasoning_effort": true,
"supports_reasoning": true,
"supports_max_reasoning_effort": true,
"supports_xhigh_reasoning_effort": true
},
"chatgpt/gpt-5.6-sol": {
@ -33843,6 +33856,7 @@
"supports_minimal_reasoning_effort": false,
"supports_none_reasoning_effort": true,
"supports_reasoning": true,
"supports_max_reasoning_effort": true,
"supports_xhigh_reasoning_effort": true
},
"chatgpt/gpt-5.6-terra": {
@ -33863,6 +33877,7 @@
"supports_minimal_reasoning_effort": false,
"supports_none_reasoning_effort": true,
"supports_reasoning": true,
"supports_max_reasoning_effort": true,
"supports_xhigh_reasoning_effort": true
},
"chatgpt/gpt-5.4": {
@ -37087,6 +37102,7 @@
"supports_prompt_cache_breakpoint": true,
"supports_prompt_caching": true,
"supports_reasoning": true,
"supports_max_reasoning_effort": true,
"supports_response_schema": true,
"supports_system_messages": true,
"supports_tool_choice": true,
@ -37162,6 +37178,7 @@
"supports_prompt_cache_breakpoint": true,
"supports_prompt_caching": true,
"supports_reasoning": true,
"supports_max_reasoning_effort": true,
"supports_response_schema": true,
"supports_system_messages": true,
"supports_tool_choice": true,
@ -37237,6 +37254,7 @@
"supports_prompt_cache_breakpoint": true,
"supports_prompt_caching": true,
"supports_reasoning": true,
"supports_max_reasoning_effort": true,
"supports_response_schema": true,
"supports_system_messages": true,
"supports_tool_choice": true,
@ -70354,7 +70372,7 @@
],
"supports_computer_use": true,
"supports_function_calling": true,
"supports_max_reasoning_effort": false,
"supports_max_reasoning_effort": true,
"supports_minimal_reasoning_effort": false,
"supports_native_streaming": true,
"supports_none_reasoning_effort": false,

View file

@ -351,16 +351,19 @@ class TestTheNormalizedTierIsTheTierSent:
("gpt-6-astra", "azure"),
("us/gpt-6-astra", "azure"),
("eu/gpt-6-astra", "azure"),
("gpt-5.6-sol", "azure"),
("gpt-5.6-terra", "azure"),
("gpt-5.6-luna", "azure"),
],
)
def test_an_azure_hosted_astra_deployment_drops_to_the_tier_its_chat_route_accepts(
self, local_model_cost_map: None, model: str, provider: str
@pytest.mark.parametrize("effort", ("max", {"effort": "max", "summary": "detailed"}))
def test_an_azure_hosted_deployment_preserves_its_declared_max_before_endpoint_selection(
self, local_model_cost_map: None, model: str, provider: str, effort: object
) -> None:
"""A Foundry gpt-6-astra answers a Chat Completions ``max`` with a 400 naming low through
xhigh while honoring the same level on the Responses route (2026-10-01). The row's flag is
route-blind, so it keeps max off and the adapter, which cannot know which route the call
takes, sends the tier below instead of the level chat rejects."""
assert _reasoning_effort_sent(model, provider, "max") == "xhigh"
"""The shared normalizer runs before the final endpoint is known. Preserve a declared
max for Responses; callers choosing plain Chat Completions remain subject to its limits."""
expected: Final = _reasoning_effort_sent("gpt-6-astra", "openai", effort)
assert _reasoning_effort_sent(model, provider, effort) == expected
def test_the_openai_hosted_twin_still_sends_max(self, local_model_cost_map):
assert _reasoning_effort_sent("gpt-6-astra", "openai", "max") == "max"

View file

@ -413,9 +413,8 @@ def test_azure_gpt6_astra_forwards_exactly_the_effort_levels_its_map_row_allows(
):
"""The azure/ row is the whole contract for a Foundry deployment. A live Foundry gpt-6-astra
(model gpt-6-astra-2026-09-03) answered reasoning_effort none with a 400 naming low, medium,
high and xhigh on 2026-10-01, so its row turns none and minimal off. The row keeps max off too,
since chat refuses it, but no gate reads that flag: the Responses route honors max, so chat
forwards it for the bridge and a plain chat call gets the provider's own 400."""
high and xhigh on 2026-10-01, so its row turns none and minimal off. Max advertises Responses
capability; chat forwards it for the bridge and a plain chat call gets the provider's own 400."""
forwarded: Final = _azure_chat_row_forwards(litellm.model_cost[model], level)
dropped: Final = config.map_openai_params(
non_default_params={"reasoning_effort": level},
@ -449,13 +448,11 @@ def test_azure_gpt6_astra_forwards_exactly_the_effort_levels_its_map_row_allows(
def test_azure_gpt6_astra_row_turns_none_and_minimal_off(model: str):
"""Pinned on purpose so the row-derived test above cannot go vacuous: a Foundry gpt-6-astra
deployment refused none on chat and on Responses on 2026-10-01 (400 unsupported_value naming
low, medium, high, xhigh), and minimal the same way; max it refused on chat while honoring it
on Responses, so the route-blind flag stays off. When Azure adds a level, flip the row and
this assertion together."""
low, medium, high, xhigh), and minimal the same way. When Azure adds either level, flip the row
and this assertion together."""
row: Final = litellm.model_cost[model]
assert row["supports_none_reasoning_effort"] is False
assert row["supports_minimal_reasoning_effort"] is False
assert row["supports_max_reasoning_effort"] is False
def test_azure_chat_gate_reads_the_foundry_row_when_azure_ai_prefix_survives_provider_remap(

View file

@ -792,14 +792,17 @@ def test_azure_responses_gpt6_effort_levels_follow_the_azure_row(local_model_cos
)
def test_azure_gpt6_astra_keeps_max_on_this_route_whatever_its_row_says(local_model_cost_map: None):
@pytest.mark.parametrize("supports_max", (False, True))
def test_azure_gpt6_astra_keeps_max_on_this_route_whatever_its_row_says(
local_model_cost_map: None, monkeypatch: pytest.MonkeyPatch, supports_max: bool
) -> None:
"""Pinned on purpose: the row-derived test above exempts max by hand, so this one has to fail
if the gate ever reads the flag. A Foundry gpt-6-astra deployment answered reasoning.effort max
through /v1/responses with 200, the effort echoed back and reasoning tokens billed, on
2026-10-01, while chat refuses the same level, as OpenAI's chat does. The row keeps max off for
the adapter and the picker, and this gate still forwards it even under drop_params."""
assert litellm.model_cost["azure/gpt-6-astra"]["supports_max_reasoning_effort"] is False
kept = AzureOpenAIResponsesAPIConfig().map_openai_params(
2026-10-01, while chat refuses the same level, as OpenAI's chat does. This gate forwards it
even under drop_params, regardless of the advertised capability."""
monkeypatch.setitem(litellm.model_cost["azure/gpt-6-astra"], "supports_max_reasoning_effort", supports_max)
kept: Final = AzureOpenAIResponsesAPIConfig().map_openai_params(
response_api_optional_params=ResponsesAPIOptionalRequestParams(reasoning={"effort": "max"}),
model="gpt-6-astra",
drop_params=True,

View file

@ -1388,8 +1388,8 @@ def test_gpt5_6_forwards_reasoning_effort_max_for_the_responses_bridge(config: O
"""A chat request carrying tools or a reasoning summary is converted to /v1/responses further
down main.py, and that surface accepts max. This runs before litellm has decided to bridge, so
refusing max here would break the cursor thinking-max shape that works today. Plain chat
completions still answer max with a provider 400, and the capability list below is what keeps
the level out of the picker."""
completions still answer max with a provider 400, so the advertised capability does not imply
support on every endpoint."""
params = config.map_openai_params(
non_default_params={"reasoning_effort": "max"},
optional_params={},
@ -1400,15 +1400,15 @@ def test_gpt5_6_forwards_reasoning_effort_max_for_the_responses_bridge(config: O
@pytest.mark.parametrize("model", ["gpt-5.6", "gpt-5.6-sol", "gpt-5.6-terra", "gpt-5.6-luna"])
def test_gpt5_6_never_advertises_reasoning_effort_max(model: str):
"""/v1/chat/completions answers max with "Unsupported value: 'reasoning_effort' does not support
'max' with this model. Supported values are: 'none', 'low', 'medium', 'high', and 'xhigh'", so no
gpt-5.6 entry asserts supports_max_reasoning_effort and the advertised set stops at xhigh."""
def test_gpt5_6_advertises_max_when_its_catalog_row_declares_it(model: str) -> None:
"""Capability discovery includes Responses-only efforts, while the provider owns endpoint
admission; the generic alias may declare a different set than named variants."""
from litellm.router_utils.reasoning_effort_capability import resolve_supported_reasoning_efforts
resolved = resolve_supported_reasoning_efforts(litellm.get_model_info(model), deployment_is_mapped=True)
model_info: Final = litellm.get_model_info(model)
resolved: Final = resolve_supported_reasoning_efforts(model_info, deployment_is_mapped=True)
assert resolved is not None
assert "max" not in resolved
assert ("max" in resolved) is (model_info["supports_max_reasoning_effort"] is True)
assert "xhigh" in resolved

View file

@ -407,27 +407,21 @@ class TestGpt6AstraAdvertisesItsDocumentedLevels:
("azure/us/gpt-6-astra", "azure"),
("azure/eu/gpt-6-astra", "azure"),
("azure_ai/gpt-6-astra", "azure_ai"),
("azure/gpt-6-astra-2026-09-03", "azure"),
],
)
def test_an_azure_hosted_deployment_advertises_the_levels_both_its_routes_accept(
self, local_model_cost_map, model, custom_llm_provider
):
"""Verified live on 2026-10-01 against a Foundry gpt-6-astra deployment (model
gpt-6-astra-2026-09-03): none returns 400 unsupported_value on chat and on Responses, naming
low, medium, high and xhigh as the levels it takes; Responses also honors max with reasoning
tokens billed while chat refuses it, as OpenAI's chat does. The flag is route-blind, so the
row keeps max off and the advertised set stops at the level every route accepts.
Earlier evidence (2026-09-05, commit e79f3ec5205) had none accepted on chat; the deployment
no longer behaves that way."""
def test_an_azure_hosted_deployment_advertises_the_same_levels_as_openai(
self, local_model_cost_map: None, model: str, custom_llm_provider: str
) -> None:
"""The flag advertises model capability; Azure documents max through Responses, so chat
rejection must not hide that capability."""
from litellm.utils import get_model_info_helper
model_info = dict(get_model_info_helper(model=model, custom_llm_provider=custom_llm_provider))
model_info: Final = dict(get_model_info_helper(model=model, custom_llm_provider=custom_llm_provider))
openai_info: Final = dict(get_model_info_helper(model="gpt-6-astra", custom_llm_provider="openai"))
assert resolve_supported_reasoning_efforts(model_info, deployment_is_mapped=True) == (
"low",
"medium",
"high",
"xhigh",
resolve_supported_reasoning_efforts(openai_info, deployment_is_mapped=True)
)

View file

@ -253,6 +253,31 @@ def test_azure_gpt_6_1_sol_rows_accept_the_same_reasoning_efforts_as_the_openai_
assert resolve_supported_reasoning_efforts(prices[row], deployment_is_mapped=True) == openai_efforts
@pytest.mark.parametrize("model", ("gpt-5.6-sol", "gpt-5.6-terra", "gpt-5.6-luna", "gpt-6-astra"))
def test_gpt_provider_rows_advertise_the_same_max_effort_as_bedrock(
prices: Mapping[str, Mapping[str, object]], model: str, local_model_cost_map: None
) -> None:
counterpart: Final = litellm.get_model_info(f"bedrock_mantle/openai.{model}")
pattern: Final = re.compile(
rf"(?:azure/(?:us/|eu/)?|azure_ai/|chatgpt/)?{re.escape(model)}(?:-\d{{4}}-\d{{2}}-\d{{2}})?"
)
rows: Final = tuple(key for key in prices if pattern.fullmatch(key))
assert rows
assert {key: prices[key].get("supports_max_reasoning_effort") for key in rows} == dict.fromkeys(
rows, counterpart["supports_max_reasoning_effort"]
)
assert {key: litellm.get_model_info(key)["supports_max_reasoning_effort"] for key in rows} == dict.fromkeys(
rows, counterpart["supports_max_reasoning_effort"]
)
assert {
key: "max"
in (resolve_supported_reasoning_efforts(litellm.get_model_info(key), deployment_is_mapped=True) or ())
for key in rows
} == dict.fromkeys(
rows, "max" in (resolve_supported_reasoning_efforts(counterpart, deployment_is_mapped=True) or ())
)
# https://ai.google.dev/gemini-api/docs/models/deep-research-preview-04-2026
@pytest.mark.parametrize("model", ["gemini/deep-research-preview-04-2026", "gemini/deep-research-max-preview-04-2026"])
def test_gemini_deep_research_previews_accept_the_documented_input_window(prices: dict, model: str):