Sorting LiteLLM GitHub issues by difficulty to implement. Higher = harder.

{
The left issue is harder because it involves provider-specific reasoning-output behavior, model-version compatibility, and validation of silent response degradation across Anthropic integrations. The right issue is a narrower request-parameter translation with a more localized implementation and test surface.
}
https://github.com/berriai/litellm/issues/25965 4:1 https://github.com/berriai/litellm/issues/27453