scriptoriumadmin/litellm
1
0
Fork 0
mirror of https://github.com/BerriAI/litellm.git synced 2026-10-10 03:28:53 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 85

refactor(cost-map): remove openrouter rows, part 1 of 5 (anthropic to google block) (#45665)

Browse source 
Co-authored-by: kerry <kerry@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
This commit is contained in:
devin-ai-integration[bot] 2026-10-09 21:51:48 +00:00 • committed by GitHub
parent aa64b07281
commit 1f1b28fc8e
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
6 changed files with 3 additions and 4728 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2345
litellm/model_prices_and_context_window_backup.json
View file

File diff suppressed because it is too large Load diff

2345
model_prices_and_context_window.json
View file

File diff suppressed because it is too large Load diff

7
tests/unit/litellm_core_utils/llm_cost_calc/test_tool_call_cost_tracking.py
View file

@ -384,15 +384,15 @@ def test_web_search_provider_prefix_fallback_does_not_misprice_non_gemini_model(
Regression for the provider-prefix fallback in _handle_web_search_cost. When the initial
get_model_info lookup fails for a "/"-containing model, the retry re-resolves model_info from
the prefix and must adopt that prefix's provider for routing. Otherwise an unrelated model
(here OpenRouter, which carries no web search pricing) is re-resolved but still routed through
(here Azure, which carries no web search pricing) is re-resolved but still routed through
the request's vertex_ai Gemini calculator, which charges its $0.035 per_prompt default for a
model that should cost nothing for web search.
"""
from litellm.types.utils import PromptTokensDetailsWrapper, Usage
model = "openrouter/google/gemini-3.1-flash-lite"
model = "azure/gpt-4o"
model_info = litellm.get_model_info(model)
assert model_info["litellm_provider"] == "openrouter"
assert model_info["litellm_provider"] == "azure"
assert not model_info.get("search_context_cost_per_query")
usage = Usage(
@ -591,4 +591,3 @@ _BEDROCK_MANTLE_WEB_SEARCH_MODELS = (
)
_BEDROCK_MANTLE_WEB_SEARCH_RATE = 0.012

24
tests/unit/litellm_core_utils/test_get_model_cost_map.py
View file

@ -212,7 +212,6 @@ def test_shipped_backup_marks_claude_4_6_plus_adaptive_not_4_0():
"anthropic.claude-opus-4-8",
"vertex_ai/claude-opus-4-6@default",
"us.anthropic.claude-sonnet-4-6",
"openrouter/anthropic/claude-opus-4.7",
"azure_ai/claude-opus-4-7",
]:
assert backup[adaptive]["supports_adaptive_thinking"] is True, adaptive
@ -224,29 +223,6 @@ def test_shipped_backup_marks_claude_4_6_plus_adaptive_not_4_0():
assert "supports_adaptive_thinking" not in backup[non_adaptive], non_adaptive
# OpenRouter headline rates from GET https://openrouter.ai/api/v1/models.
# These were the catalog values that disagreed with that API (and, for the
# two spotlight models, the public model pages that their source fields cite).
_OPENROUTER_LIVE_COSTS = {
"openrouter/qwen/qwen3.5-plus-02-15": (2.6e-07, 1.56e-06, None),
"openrouter/openai/gpt-oss-120b": (3.7e-08, 1.7e-07, None),
"openrouter/qwen/qwen3-coder-plus": (6.5e-07, 3.25e-06, None),
"openrouter/qwen/qwen3.5-flash-02-23": (6.5e-08, 2.6e-07, None),
"openrouter/qwen/qwen3.5-27b": (1.95e-07, 1.56e-06, None),
"openrouter/gryphe/mythomax-l2-13b": (6e-08, 6e-08, None),
"openrouter/mancer/weaver": (4e-07, 7.5e-07, None),
"openrouter/xiaomi/mimo-v2.5-pro": (4.35e-07, 8.7e-07, 3.6e-09),
"openrouter/moonshotai/kimi-k2.5": (4.5e-07, 2.25e-06, 7e-08),
"openrouter/z-ai/glm-5": (6e-07, 1.92e-06, None),
}
_OPENROUTER_STALE_COSTS = {
"openrouter/qwen/qwen3.5-plus-02-15": (4e-07, 2.4e-06),
"openrouter/openai/gpt-oss-120b": (1.8e-07, 8e-07),
"openrouter/gryphe/mythomax-l2-13b": (1.875e-06, 1.875e-06),
}
def test_get_model_cost_map_stamps_loaded_at():
"""The load time feeds each pod's reload-due decision; a load that does not stamp it
would make manual reload requests race the proxy's startup"""

1
tests/unit/llms/anthropic/chat/test_anthropic_chat_transformation.py
View file

@ -2787,7 +2787,6 @@ def test_raw_adaptive_thinking_untouched_for_46_plus_model():
("claude-opus-4-8", True),
("anthropic.claude-opus-4-8", True),
("vertex_ai/claude-opus-4-6@default", True),
("openrouter/anthropic/claude-opus-4.7", True),
("us.anthropic.claude-sonnet-4-6", True),
("claude-opus-4-6-20260205", True),
# unmapped future models -> anthropic-claude fallback rule

9
tests/unit/test_utils.py
View file

@ -4144,15 +4144,6 @@ ANTHROPIC_REEXPORT_CACHE_MIN: Final = {
"databricks/databricks-claude-sonnet-4": 1024,
"databricks/databricks-claude-sonnet-4-5": 1024,
"databricks/databricks-claude-sonnet-4-6": 1024,
"openrouter/anthropic/claude-haiku-4.5": 4096,
"openrouter/anthropic/claude-opus-4": 1024,
"openrouter/anthropic/claude-opus-4.1": 1024,
"openrouter/anthropic/claude-opus-4.5": 4096,
"openrouter/anthropic/claude-opus-4.6": 4096,
"openrouter/anthropic/claude-opus-4.7": 2048,
"openrouter/anthropic/claude-sonnet-4": 1024,
"openrouter/anthropic/claude-sonnet-4.5": 1024,
"openrouter/anthropic/claude-sonnet-4.6": 1024,
"replicate/anthropic/claude-4-sonnet": 1024,
"replicate/anthropic/claude-4.5-haiku": 4096,
"replicate/anthropic/claude-4.5-sonnet": 1024,