PR #28229: fix: use router deployment id for anthropic messages cost

• Repository: BerriAI/litellm
• Author: kongjiadongyuan
• State: open | merged: False
• Link: https://github.com/BerriAI/litellm/pull/28229

Description (problem / solution / changelog)

Relevant issues

Fixes #28228

Linear ticket

N/A

Pre-Submission checklist

Please complete all items before asking a LiteLLM maintainer to review your PR

• I have Added testing in the tests/test_litellm/ directory, Adding at least 1 test is a hard requirement - see details
• My PR passes all unit tests on make test-unit
  • Not run locally; scoped tests below were run instead.
• My PR's scope is as isolated as possible, it only solves 1 specific problem
• I have requested a Greptile review by commenting @greptileai and received a Confidence Score of at least 4/5 before requesting a maintainer review

Delays in PR merge?

If you're seeing a delay in your PR being merged, ping the LiteLLM Team on Slack (#pr-review).

CI (LiteLLM team)

CI status guideline:

• 50-55 passing tests: main is stable with minor issues.
• 45-49 passing tests: acceptable but needs attention
• <= 40 passing tests: unstable; be careful with your merges and assess the risk.

• Branch creation CI run
  Link:

• CI run for the last commit
  Link:

• Merge / cherry-pick CI run
  Links:

Screenshots / Proof of Fix

Scoped local verification:

python3 -m py_compile litellm/litellm_core_utils/litellm_logging.py litellm/proxy/pass_through_endpoints/llm_provider_handlers/anthropic_passthrough_logging_handler.py litellm/proxy/common_request_processing.py tests/test_litellm/litellm_core_utils/test_litellm_logging.py tests/test_litellm/proxy/test_common_request_processing.py
# passed

git diff --check
# passed

uv run pytest tests/test_litellm/litellm_core_utils/test_litellm_logging.py -k 'router_model_id or custom_pricing or AnthropicPassthroughCustomPricing or direct_model_info' -q
# 9 passed, 73 deselected

uv run --extra proxy pytest tests/test_litellm/proxy/test_common_request_processing.py -k 'response_cost or get_custom_headers' -q
# 8 passed, 80 deselected

uv run black --check litellm/litellm_core_utils/litellm_logging.py litellm/proxy/common_request_processing.py litellm/proxy/pass_through_endpoints/llm_provider_handlers/anthropic_passthrough_logging_handler.py tests/test_litellm/litellm_core_utils/test_litellm_logging.py tests/test_litellm/proxy/test_common_request_processing.py
# passed

uv run ruff check litellm/litellm_core_utils/litellm_logging.py litellm/proxy/common_request_processing.py litellm/proxy/pass_through_endpoints/llm_provider_handlers/anthropic_passthrough_logging_handler.py
# passed

uv run ruff check over the full touched test files was not used as a pass/fail signal because tests/test_litellm/litellm_core_utils/test_litellm_logging.py currently has pre-existing lint violations unrelated to this change.

Type

🐛 Bug Fix ✅ Test

Changes

This fixes Anthropic native /v1/messages cost tracking when several router deployments share the same public model group but have deployment-specific custom pricing.

• Prefer direct litellm_params["model_info"].id when resolving the router deployment id for cost lookup, before falling back to litellm_metadata / metadata.
• Detect custom pricing in direct litellm_params["model_info"], not only nested metadata.
• Preserve calculated response cost in hidden params after the logging cost calculator runs, so proxy headers/spend can use the same cost.
• Pass the logging object into Anthropic passthrough completion_cost() and store the calculated response cost on model_call_details, matching other passthrough handlers.
• Fall back to litellm_logging_obj.model_call_details["response_cost"] in proxy custom headers when response hidden params do not contain a cost yet.
• Add regression coverage for two deployments under one public model group with different custom pricing, verifying cost uses the selected deployment id.

Changed files

• litellm/litellm_core_utils/litellm_logging.py (modified, +22/-3)
• litellm/proxy/common_request_processing.py (modified, +7/-0)
• litellm/proxy/pass_through_endpoints/llm_provider_handlers/anthropic_passthrough_logging_handler.py (modified, +2/-0)
• tests/test_litellm/litellm_core_utils/test_litellm_logging.py (modified, +106/-0)
• tests/test_litellm/proxy/test_common_request_processing.py (modified, +36/-0)