Five Keys to Controlling AI Token Costs
Model routing, semantic caching, prompt caching, reranking and response constraints are the five keys InfoWorld names for controlling AI token costs. We price each one at October 2026 list rates from Anthropic, OpenAI and Google, flag what changed in prompt caching, and work through a support-bot example that takes a modelled $62,800 monthly bill to about $7,213.
