scriptoriumadmin/litellm
1
0
Fork 0
mirror of https://github.com/BerriAI/litellm.git synced 2026-09-08 22:21:35 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 78

Merge 0400f0b739 into 40423e6ec0

Browse source
This commit is contained in:
jlan-nl 2026-08-26 08:02:37 -04:00 • committed by GitHub
commit 59f3da97e1
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
3 changed files with 123 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

36
litellm/model_prices_and_context_window_backup.json
View file

@ -6592,6 +6592,9 @@
"supports_web_search": true
},
"azure/gpt-5.6": {
"cache_creation_input_token_cost": 6.25e-06,
"cache_creation_input_token_cost_above_272k_tokens": 1.25e-05,
"cache_creation_input_token_cost_priority": 1.25e-05,
"cache_read_input_token_cost": 5e-07,
"cache_read_input_token_cost_above_272k_tokens": 1e-06,
"cache_read_input_token_cost_priority": 1e-06,
@ -6642,6 +6645,9 @@
"supports_minimal_reasoning_effort": false
},
"azure/gpt-5.6-sol": {
"cache_creation_input_token_cost": 6.25e-06,
"cache_creation_input_token_cost_above_272k_tokens": 1.25e-05,
"cache_creation_input_token_cost_priority": 1.25e-05,
"cache_read_input_token_cost": 5e-07,
"cache_read_input_token_cost_above_272k_tokens": 1e-06,
"cache_read_input_token_cost_priority": 1e-06,
@ -6693,6 +6699,9 @@
"supports_minimal_reasoning_effort": false
},
"azure/gpt-5.6-terra": {
"cache_creation_input_token_cost": 2.5e-06,
"cache_creation_input_token_cost_above_272k_tokens": 5e-06,
"cache_creation_input_token_cost_priority": 5e-06,
"cache_read_input_token_cost": 2e-07,
"cache_read_input_token_cost_above_272k_tokens": 4e-07,
"cache_read_input_token_cost_priority": 4e-07,
@ -6744,6 +6753,9 @@
"supports_minimal_reasoning_effort": false
},
"azure/gpt-5.6-luna": {
"cache_creation_input_token_cost": 2.5e-07,
"cache_creation_input_token_cost_above_272k_tokens": 5e-07,
"cache_creation_input_token_cost_priority": 5e-07,
"cache_read_input_token_cost": 2e-08,
"cache_read_input_token_cost_above_272k_tokens": 4e-08,
"cache_read_input_token_cost_priority": 4e-08,
@ -6795,6 +6807,9 @@
"supports_minimal_reasoning_effort": false
},
"azure/us/gpt-5.6": {
"cache_creation_input_token_cost": 6.875e-06,
"cache_creation_input_token_cost_above_272k_tokens": 1.375e-05,
"cache_creation_input_token_cost_priority": 1.71875e-05,
"cache_read_input_token_cost": 5.5e-07,
"cache_read_input_token_cost_above_272k_tokens": 1.1e-06,
"cache_read_input_token_cost_priority": 1.375e-06,
@ -6842,6 +6857,9 @@
"supports_minimal_reasoning_effort": false
},
"azure/us/gpt-5.6-sol": {
"cache_creation_input_token_cost": 6.875e-06,
"cache_creation_input_token_cost_above_272k_tokens": 1.375e-05,
"cache_creation_input_token_cost_priority": 1.71875e-05,
"cache_read_input_token_cost": 5.5e-07,
"cache_read_input_token_cost_above_272k_tokens": 1.1e-06,
"cache_read_input_token_cost_priority": 1.375e-06,
@ -6890,6 +6908,9 @@
"supports_minimal_reasoning_effort": false
},
"azure/us/gpt-5.6-terra": {
"cache_creation_input_token_cost": 2.75e-06,
"cache_creation_input_token_cost_above_272k_tokens": 5.5e-06,
"cache_creation_input_token_cost_priority": 6.875e-06,
"cache_read_input_token_cost": 2.2e-07,
"cache_read_input_token_cost_above_272k_tokens": 4.4e-07,
"cache_read_input_token_cost_priority": 5.5e-07,
@ -6938,6 +6959,9 @@
"supports_minimal_reasoning_effort": false
},
"azure/us/gpt-5.6-luna": {
"cache_creation_input_token_cost": 2.75e-07,
"cache_creation_input_token_cost_above_272k_tokens": 5.5e-07,
"cache_creation_input_token_cost_priority": 6.875e-07,
"cache_read_input_token_cost": 2.2e-08,
"cache_read_input_token_cost_above_272k_tokens": 4.4e-08,
"cache_read_input_token_cost_priority": 5.5e-08,
@ -6986,6 +7010,9 @@
"supports_minimal_reasoning_effort": false
},
"azure/eu/gpt-5.6": {
"cache_creation_input_token_cost": 6.875e-06,
"cache_creation_input_token_cost_above_272k_tokens": 1.375e-05,
"cache_creation_input_token_cost_priority": 1.71875e-05,
"cache_read_input_token_cost": 5.5e-07,
"cache_read_input_token_cost_above_272k_tokens": 1.1e-06,
"cache_read_input_token_cost_priority": 1.375e-06,
@ -7033,6 +7060,9 @@
"supports_minimal_reasoning_effort": false
},
"azure/eu/gpt-5.6-sol": {
"cache_creation_input_token_cost": 6.875e-06,
"cache_creation_input_token_cost_above_272k_tokens": 1.375e-05,
"cache_creation_input_token_cost_priority": 1.71875e-05,
"cache_read_input_token_cost": 5.5e-07,
"cache_read_input_token_cost_above_272k_tokens": 1.1e-06,
"cache_read_input_token_cost_priority": 1.375e-06,
@ -7081,6 +7111,9 @@
"supports_minimal_reasoning_effort": false
},
"azure/eu/gpt-5.6-terra": {
"cache_creation_input_token_cost": 2.75e-06,
"cache_creation_input_token_cost_above_272k_tokens": 5.5e-06,
"cache_creation_input_token_cost_priority": 6.875e-06,
"cache_read_input_token_cost": 2.2e-07,
"cache_read_input_token_cost_above_272k_tokens": 4.4e-07,
"cache_read_input_token_cost_priority": 5.5e-07,
@ -7129,6 +7162,9 @@
"supports_minimal_reasoning_effort": false
},
"azure/eu/gpt-5.6-luna": {
"cache_creation_input_token_cost": 2.75e-07,
"cache_creation_input_token_cost_above_272k_tokens": 5.5e-07,
"cache_creation_input_token_cost_priority": 6.875e-07,
"cache_read_input_token_cost": 2.2e-08,
"cache_read_input_token_cost_above_272k_tokens": 4.4e-08,
"cache_read_input_token_cost_priority": 5.5e-08,

36
model_prices_and_context_window.json
View file

@ -6592,6 +6592,9 @@
"supports_web_search": true
},
"azure/gpt-5.6": {
"cache_creation_input_token_cost": 6.25e-06,
"cache_creation_input_token_cost_above_272k_tokens": 1.25e-05,
"cache_creation_input_token_cost_priority": 1.25e-05,
"cache_read_input_token_cost": 5e-07,
"cache_read_input_token_cost_above_272k_tokens": 1e-06,
"cache_read_input_token_cost_priority": 1e-06,
@ -6642,6 +6645,9 @@
"supports_minimal_reasoning_effort": false
},
"azure/gpt-5.6-sol": {
"cache_creation_input_token_cost": 6.25e-06,
"cache_creation_input_token_cost_above_272k_tokens": 1.25e-05,
"cache_creation_input_token_cost_priority": 1.25e-05,
"cache_read_input_token_cost": 5e-07,
"cache_read_input_token_cost_above_272k_tokens": 1e-06,
"cache_read_input_token_cost_priority": 1e-06,
@ -6693,6 +6699,9 @@
"supports_minimal_reasoning_effort": false
},
"azure/gpt-5.6-terra": {
"cache_creation_input_token_cost": 2.5e-06,
"cache_creation_input_token_cost_above_272k_tokens": 5e-06,
"cache_creation_input_token_cost_priority": 5e-06,
"cache_read_input_token_cost": 2e-07,
"cache_read_input_token_cost_above_272k_tokens": 4e-07,
"cache_read_input_token_cost_priority": 4e-07,
@ -6744,6 +6753,9 @@
"supports_minimal_reasoning_effort": false
},
"azure/gpt-5.6-luna": {
"cache_creation_input_token_cost": 2.5e-07,
"cache_creation_input_token_cost_above_272k_tokens": 5e-07,
"cache_creation_input_token_cost_priority": 5e-07,
"cache_read_input_token_cost": 2e-08,
"cache_read_input_token_cost_above_272k_tokens": 4e-08,
"cache_read_input_token_cost_priority": 4e-08,
@ -6795,6 +6807,9 @@
"supports_minimal_reasoning_effort": false
},
"azure/us/gpt-5.6": {
"cache_creation_input_token_cost": 6.875e-06,
"cache_creation_input_token_cost_above_272k_tokens": 1.375e-05,
"cache_creation_input_token_cost_priority": 1.71875e-05,
"cache_read_input_token_cost": 5.5e-07,
"cache_read_input_token_cost_above_272k_tokens": 1.1e-06,
"cache_read_input_token_cost_priority": 1.375e-06,
@ -6842,6 +6857,9 @@
"supports_minimal_reasoning_effort": false
},
"azure/us/gpt-5.6-sol": {
"cache_creation_input_token_cost": 6.875e-06,
"cache_creation_input_token_cost_above_272k_tokens": 1.375e-05,
"cache_creation_input_token_cost_priority": 1.71875e-05,
"cache_read_input_token_cost": 5.5e-07,
"cache_read_input_token_cost_above_272k_tokens": 1.1e-06,
"cache_read_input_token_cost_priority": 1.375e-06,
@ -6890,6 +6908,9 @@
"supports_minimal_reasoning_effort": false
},
"azure/us/gpt-5.6-terra": {
"cache_creation_input_token_cost": 2.75e-06,
"cache_creation_input_token_cost_above_272k_tokens": 5.5e-06,
"cache_creation_input_token_cost_priority": 6.875e-06,
"cache_read_input_token_cost": 2.2e-07,
"cache_read_input_token_cost_above_272k_tokens": 4.4e-07,
"cache_read_input_token_cost_priority": 5.5e-07,
@ -6938,6 +6959,9 @@
"supports_minimal_reasoning_effort": false
},
"azure/us/gpt-5.6-luna": {
"cache_creation_input_token_cost": 2.75e-07,
"cache_creation_input_token_cost_above_272k_tokens": 5.5e-07,
"cache_creation_input_token_cost_priority": 6.875e-07,
"cache_read_input_token_cost": 2.2e-08,
"cache_read_input_token_cost_above_272k_tokens": 4.4e-08,
"cache_read_input_token_cost_priority": 5.5e-08,
@ -6986,6 +7010,9 @@
"supports_minimal_reasoning_effort": false
},
"azure/eu/gpt-5.6": {
"cache_creation_input_token_cost": 6.875e-06,
"cache_creation_input_token_cost_above_272k_tokens": 1.375e-05,
"cache_creation_input_token_cost_priority": 1.71875e-05,
"cache_read_input_token_cost": 5.5e-07,
"cache_read_input_token_cost_above_272k_tokens": 1.1e-06,
"cache_read_input_token_cost_priority": 1.375e-06,
@ -7033,6 +7060,9 @@
"supports_minimal_reasoning_effort": false
},
"azure/eu/gpt-5.6-sol": {
"cache_creation_input_token_cost": 6.875e-06,
"cache_creation_input_token_cost_above_272k_tokens": 1.375e-05,
"cache_creation_input_token_cost_priority": 1.71875e-05,
"cache_read_input_token_cost": 5.5e-07,
"cache_read_input_token_cost_above_272k_tokens": 1.1e-06,
"cache_read_input_token_cost_priority": 1.375e-06,
@ -7081,6 +7111,9 @@
"supports_minimal_reasoning_effort": false
},
"azure/eu/gpt-5.6-terra": {
"cache_creation_input_token_cost": 2.75e-06,
"cache_creation_input_token_cost_above_272k_tokens": 5.5e-06,
"cache_creation_input_token_cost_priority": 6.875e-06,
"cache_read_input_token_cost": 2.2e-07,
"cache_read_input_token_cost_above_272k_tokens": 4.4e-07,
"cache_read_input_token_cost_priority": 5.5e-07,
@ -7129,6 +7162,9 @@
"supports_minimal_reasoning_effort": false
},
"azure/eu/gpt-5.6-luna": {
"cache_creation_input_token_cost": 2.75e-07,
"cache_creation_input_token_cost_above_272k_tokens": 5.5e-07,
"cache_creation_input_token_cost_priority": 6.875e-07,
"cache_read_input_token_cost": 2.2e-08,
"cache_read_input_token_cost_above_272k_tokens": 4.4e-08,
"cache_read_input_token_cost_priority": 5.5e-08,

51
tests/test_litellm/test_cost_calculator.py
View file

@ -1,4 +1,6 @@
from typing import Final
import pytest
@ -3781,3 +3783,52 @@ def test_completion_cost_prices_anthropic_shaped_cache_read_tokens(_local_model_
)
assert cost == pytest.approx(3 * 4e-6 + 4014 * 4e-7 + 5 * 2e-5, rel=1e-9)
@pytest.mark.parametrize(
"model",
tuple(
f"azure/{region}gpt-5.6{variant}"
for region in ("", "eu/", "us/")
for variant in ("", "-sol", "-terra", "-luna")
),
)
@pytest.mark.parametrize("tier", ("", "_above_272k_tokens", "_priority"))
def test_azure_gpt_5_6_cache_write_rate_matches_openai_ratio(_local_model_cost_map, model: str, tier: str):
"""Regression for #37631: the azure/* gpt-5.6 entries carried no
cache_creation_input_token_cost at all, so cache writes fell back to $0. Azure
charges 1.25x the input rate for them, matching the OpenAI-hosted twins."""
info: Final = litellm.model_cost[model]
twin: Final = litellm.model_cost[model.split("/")[-1]]
twin_ratio: Final = twin[f"cache_creation_input_token_cost{tier}"] / twin[f"input_cost_per_token{tier}"]
assert twin_ratio == pytest.approx(1.25, rel=1e-12)
assert info[f"cache_creation_input_token_cost{tier}"] == pytest.approx(
twin_ratio * info[f"input_cost_per_token{tier}"], rel=1e-12
)
def test_azure_gpt_5_6_bills_cache_write_tokens_at_cache_creation_rate(_local_model_cost_map):
"""Regression for #37631: Azure reports cache writes as
prompt_tokens_details.cache_write_tokens, which LiteLLM classifies as cache-creation
tokens. Without a cache-creation rate on the azure/* entries the lookup yielded $0,
under-billing a write-heavy prompt by ~98%."""
from litellm.litellm_core_utils.llm_cost_calc.utils import generic_cost_per_token
def _input_cost(model: str) -> float:
usage: Final = Usage(
completion_tokens=0,
prompt_tokens=10_000,
total_tokens=10_000,
prompt_tokens_details=PromptTokensDetailsWrapper(
cached_tokens=0, text_tokens=0, cache_write_tokens=10_000
),
)
return generic_cost_per_token(model=model, usage=usage, custom_llm_provider="azure")[0]
global_cost: Final = _input_cost("azure/gpt-5.6-sol")
eu_cost: Final = _input_cost("azure/eu/gpt-5.6-sol")
assert global_cost == pytest.approx(10_000 * 6.25e-06, rel=1e-9)
assert global_cost > 10_000 * litellm.model_cost["azure/gpt-5.6-sol"]["input_cost_per_token"]
assert eu_cost == pytest.approx(1.1 * global_cost, rel=1e-9)