Mehmet Beydogan developed a token-based cost model for the Kimi-K2 Thinking Model within the Helicone/helicone repository, focusing on improving cost visibility and budgeting accuracy for model usage. He designed and implemented a system in TypeScript that calculates prompt and completion token costs, enabling more granular tracking and data-driven pricing decisions. The work established a foundation for future cost analytics and usage reporting, with clear traceability to the relevant commit and ticket. Mehmet’s approach demonstrated depth in cost modeling and TypeScript development, addressing business needs for transparency and control without introducing major bugs, and providing a maintainable solution for ongoing enhancements.