From eed760a37d91d462bfaa5e8c068ceadf51d21bae Mon Sep 17 00:00:00 2001 From: kerry Date: Wed, 16 Sep 2026 16:49:15 +0000 Subject: [PATCH] test: drop azure gpt-5.6 rate tests pinned to a dated price page Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- tests/test_litellm/test_cost_calculator.py | 67 ---------------------- 1 file changed, 67 deletions(-) diff --git a/tests/test_litellm/test_cost_calculator.py b/tests/test_litellm/test_cost_calculator.py index 7b53d3a58df..a5ed7175649 100644 --- a/tests/test_litellm/test_cost_calculator.py +++ b/tests/test_litellm/test_cost_calculator.py @@ -1616,73 +1616,6 @@ def test_azure_ai_cache_cost_calculation(_local_model_cost_map): ) -AZURE_GPT_5_6_MAP_KEYS = ( - "azure/gpt-5.6", - "azure/gpt-5.6-sol", - "azure/gpt-5.6-terra", - "azure/gpt-5.6-luna", - "azure/us/gpt-5.6", - "azure/us/gpt-5.6-sol", - "azure/us/gpt-5.6-terra", - "azure/us/gpt-5.6-luna", - "azure/eu/gpt-5.6", - "azure/eu/gpt-5.6-sol", - "azure/eu/gpt-5.6-terra", - "azure/eu/gpt-5.6-luna", -) - - -def test_azure_gpt_5_6_cache_write_tokens_are_billed(_local_model_cost_map): - """ - Azure bills gpt-5.6 prompt cache writes at 1.25x the input rate on every - tier, but the azure entries carried no ``cache_creation_input_token_cost``, - so cache-write tokens were billed at the plain input rate instead. - """ - from litellm.litellm_core_utils.llm_cost_calc.utils import generic_cost_per_token - from litellm.types.utils import PromptTokensDetailsWrapper, Usage - - usage = Usage( - completion_tokens=100, - prompt_tokens=2000, - total_tokens=2100, - prompt_tokens_details=PromptTokensDetailsWrapper(cached_tokens=0, text_tokens=687), - cache_creation_input_tokens=1313, - ) - - input_cost, output_cost = generic_cost_per_token( - model="azure/gpt-5.6-luna", usage=usage, custom_llm_provider="azure" - ) - - assert input_cost == pytest.approx(687 * 2e-07 + 1313 * 2.5e-07) - assert output_cost == pytest.approx(100 * 1.2e-06) - - -@pytest.mark.parametrize("model", AZURE_GPT_5_6_MAP_KEYS) -def test_azure_gpt_5_6_rates_match_azure_price_page(_local_model_cost_map, model): - """ - Per the Azure OpenAI price page (rendered 2026-08-26): cache writes cost - 1.25x input on every gpt-5.6 tier, and Data Zone costs 1.1x Global for - standard and priority alike (us/eu priority rates previously sat at 1.25x). - """ - entry = litellm.model_cost[model] - input_keys = [key for key in entry if key.startswith("input_cost_per_token")] - assert input_keys - for key in input_keys: - suffix = key[len("input_cost_per_token") :] - assert entry["cache_creation_input_token_cost" + suffix] == pytest.approx(entry[key] * 1.25) - - zone = model.split("/")[1] - if zone in ("us", "eu"): - global_entry = litellm.model_cost["azure/" + model.split("/", 2)[2]] - prefixes = ("input_cost_per_token", "output_cost_per_token", "cache_read", "cache_creation") - token_cost_keys = [key for key in entry if key.startswith(prefixes)] - global_token_cost_keys = [key for key in global_entry if key.startswith(prefixes)] - assert len(token_cost_keys) >= 9 - assert sorted(token_cost_keys) == sorted(global_token_cost_keys) - for key in token_cost_keys: - assert entry[key] == pytest.approx(global_entry[key] * 1.1), key - - def test_vertex_regional_deployment_costs_uplift_over_global(monkeypatch): """ Regression for https://github.com/BerriAI/litellm/issues/34393: two Vertex