Worked on the BerriAI/litellm repository to update pricing logic for GPT-4o-Transcribe and GPT-4o-Mini-Transcribe models, ensuring reduced per-audio-token costs were accurately reflected in the backend. Addressed a pricing calculation bug to improve cost estimation reliability within transcribe workflows, supporting better budgeting for users. The implementation involved updating Python code and JSON-based configurations, as well as enhancing unit tests to validate the new pricing structure. Emphasized API integration and backend development practices to maintain code quality and traceability, resulting in more predictable and transparent cost calculations for both customers and internal stakeholders during the monthly development cycle.