litellm - 💡(How to fix) Fix [Feature]: cost_per_token() in cost_calculator.py should return (0.0, 0.0) for unrecognized models instead of raising [3 comments, 2 participants]

Check for existing issues

• I have searched the existing issues and checked that my issue is not a duplicate.

The Feature

cost_per_token() and completion_cost() in litellm/cost_calculator.py raise an exception when the model is not found in the pricing database:

from litellm import cost_per_token

cost_per_token(
    model="openrouter/meta-llama/llama-3.3-70b-instruct",
    completion_tokens=50,
)   
# → raises: No cost information for model

Proposed fix — return (0.0, 0.0) with a UserWarning instead of raising:

# in cost_calculator.py
warnings.warn(f"No pricing data for model '{model}'. Returning (0.0, 0.0).")
return 0.0, 0.0

Motivation, pitch

Raising breaks any flow iterating over a mixed list of models. Downstream tools using LiteLLM's pricing functions are forced to wrap every single call in try/except and implement their own fallbacks — which defeats the purpose of having a shared pricing utility.

Returning (0.0, 0.0) with a warning is non-breaking: callers that already handle zero cost are unaffected, and callers that relied on the exception can check for zero instead.

Context

Ran into this while building Baar-Core, a pre-flight budget kill-switch for LLM agents that uses LiteLLM's cost_per_token() and completion_cost() for local cost enforcement — no proxy required.

This covers a gap that LiteLLM's existing budget manager doesn't address: SDK-side pre-flight enforcement for users not running the proxy. OpenRouter users hit this exception on their very first call.

Happy to open a PR for the fix in cost_calculator.py if this direction works for the maintainers. If the team is open to it, Baar-Core could also be listed as a companion library for SDK users who want hard budget limits without spinning up the proxy.

What part of LiteLLM is this about?

SDK (litellm Python package)

LiteLLM is hiring a founding backend engineer, are you interested in joining us and shipping to all our users?

Yes

Twitter / LinkedIn details

@orvi_onethread / https://bd.linkedin.com/in/robatdasorvi-productdev