fix(router): propagate custom cost_per_token from db model_info in fallback path

[Zerohertz]

Relevant issues

Fixes #25874

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
• 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

Before	After

Unit tests — 2 new tests added and passing:

• test_model_group_info_cost_from_db_model_info — verifies input_cost_per_token / output_cost_per_token are read from db_model_info when get_deployment_model_info fails
• test_model_group_info_cost_none_when_db_model_info_has_no_cost — verifies values remain None when db_model_info has no cost fields

tests/test_litellm/test_router.py::test_model_group_info_cost_from_db_model_info PASSED
tests/test_litellm/test_router.py::test_model_group_info_cost_none_when_db_model_info_has_no_cost PASSED

Type

🐛 Bug Fix

Changes

When a model is not found in LiteLLM's built-in model cost map (model_info is None in get_model_group_info()), the fallback ModelMapInfo was hardcoding input_cost_per_token and output_cost_per_token to None, ignoring any custom cost values the user had configured in the database's model_info.

This fix reads input_cost_per_token and output_cost_per_token from db_model_info (the user's database/config model_info), consistent with how mode is already read from the same source. If the keys are not present, .get() returns None, preserving backward compatibility.

Before

model_info = ModelMapInfo(
    ...
    input_cost_per_token=None,
    output_cost_per_token=None,
    ...
)

After

input_cost_per_token = db_model_info.get("input_cost_per_token")
output_cost_per_token = db_model_info.get("output_cost_per_token")

model_info = ModelMapInfo(
    ...
    input_cost_per_token=input_cost_per_token,
    output_cost_per_token=output_cost_per_token,
    ...
)

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
litellm	Ready	Preview, Comment	Apr 16, 2026 6:57pm

[CLAassistant]

All committers have signed the CLA.

[greptile-apps]

Greptile Summary

Fixes a bug in Router.get_model_group_info() where custom input_cost_per_token / output_cost_per_token values stored in the DB/config model_info were silently ignored when a model was not found in LiteLLM's built-in cost map. The fix mirrors the pre-existing mode extraction from db_model_info and preserves backward compatibility via .get() returning None when the keys are absent.

Confidence Score: 5/5

Safe to merge — the fix is minimal, backward-compatible, and well-tested.

The production change is two lines that mirror an existing pattern (mode extraction) and are guarded by .get() for backward compatibility. Both new tests are mock-only and pass. The only finding is a P2 style issue (redundant inline import) that does not affect correctness.

No files require special attention.

Important Files Changed

Filename	Overview
litellm/router.py	Reads input_cost_per_token and output_cost_per_token from db_model_info in the model_info is None fallback branch — consistent with the existing mode extraction pattern; no logic concerns.
tests/test_litellm/test_router.py	Adds two unit tests covering the new fallback cost propagation; tests are mock-only (no real network calls). Minor: from unittest.mock import patch is imported inline inside both new test functions, but patch is already at module level on line 5.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[get_model_group_info] --> B{model_id present?}
    B -- yes --> C[get_deployment_model_info]
    C -- success --> D[model_info from LiteLLM map]
    C -- exception --> E[model_info = None fallback]
    B -- no --> E
    E --> F[db_model_info = model.get model_info]
    F --> G[Extract mode]
    F --> H[Extract input cost per token - NEW]
    F --> I[Extract output cost per token - NEW]
    G --> J[Build ModelMapInfo fallback]
    H --> J
    I --> J
    D --> K[Use model_info directly]
    J --> L[Build ModelGroupInfo]
    K --> L

Loading

_{Reviews (1): Last reviewed commit: "test(router): propagate custom cost_per_..." | Re-trigger Greptile}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[Zerohertz]

There are over 300 Linting CI issues.
I haven't taken any action because this isn't related to this PR, but please reply if necessary!

334 files would be reformatted, 1411 files would be left unchanged.