About this session

It is Summer 2026 and the world is burning for token consumption—figuratively and literally. Accelerating frontier model capabilities increasingly allow agents to operate across long-running, highly parallelized tasks at exponential inference growth. In this talk, I explain how dynamic model routing—intelligently directing agent requests to the best-suited model at the best price—can reduce token costs by up to 90% while maintaining or improving accuracy. I walk through how routing works, when it doesn't, and why the world (and your agent) need routing to scale intelligence to infinity and beyond.

Topics

Speaker

Tomás Hernando Kofman

Notdiamond