scriptoriumadmin/litellm
1
0
Fork 0
mirror of https://github.com/BerriAI/litellm.git synced 2026-10-27 06:13:07 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 72

fix(cost): preserve rates across mixed token thresholds

Browse source
This commit is contained in:
Joshua Valluru 2026-10-19 22:40:36 -07:00
parent 4e6b2e1d3d
commit ee51bd07e5
2 changed files with 31 additions and 3 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

4
litellm/litellm_core_utils/llm_cost_calc/utils.py
View file

@ -696,7 +696,7 @@ def _get_token_base_cost(
# Only sort the threshold keys (typically 1-2 keys instead of 66+)
threshold: float | None = None
for key in sorted(threshold_keys, key=_parse_above_token_threshold, reverse=True):
for key in sorted(threshold_keys, key=_parse_above_token_threshold):
value = model_info.get(key)
if value is not None:
try:
@ -770,8 +770,6 @@ def _get_token_base_cost(
)
cache_read_cost = get_cost_per_unit(model_info, cache_read_tiered_key, cache_read_cost)
break
except (IndexError, ValueError):
continue
except Exception:

30
tests/unit/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py
View file

@ -2550,6 +2550,36 @@ def test_flat_input_price_keeps_other_token_thresholds(field: str) -> None:
assert rates[4] == (0.005 if field == "cache_read_input_token_cost" else 0.003)
@pytest.mark.parametrize("prompt_tokens", (200_000, 200_001, 512_000, 512_001))
@pytest.mark.parametrize("higher_rate", (0.0, 0.009))
def test_mixed_token_thresholds_preserve_each_crossed_rate(prompt_tokens: int, higher_rate: float) -> None:
from typing import Final, cast
from litellm.litellm_core_utils.llm_cost_calc.utils import _get_token_base_cost
info: Final = cast(
ModelInfo,
{
"input_cost_per_token": 0.001,
"output_cost_per_token": 0.002,
"cache_read_input_token_cost": 0.003,
"cache_creation_input_token_cost": 0.004,
"input_cost_per_token_above_200k_tokens": 0.005,
"output_cost_per_token_above_200k_tokens": 0.006,
"cache_creation_input_token_cost_above_200k_tokens": 0.007,
"cache_read_input_token_cost_above_512k_tokens": higher_rate,
"output_cost_per_token_above_512k_tokens": higher_rate,
},
)
rates: Final = _get_token_base_cost(
model_info=info,
usage=Usage(prompt_tokens=prompt_tokens, completion_tokens=3, total_tokens=prompt_tokens + 3),
)
assert rates[0] == (0.005 if prompt_tokens > 200_000 else 0.001)
assert rates[1] == (higher_rate if prompt_tokens > 512_000 else 0.006 if prompt_tokens > 200_000 else 0.002)
assert rates[2] == (0.007 if prompt_tokens > 200_000 else 0.004)
assert rates[4] == (higher_rate if prompt_tokens > 512_000 else 0.003)
def test_threshold_keys_exclude_service_tier_variants():
from typing import cast