Tools
As agents use more context, input tokens have become the majority of price-equivalent token costs.
As AI agents handle increasingly complex tasks with extended context windows, the cost structure of LLM usage has shifted such that input tokens now represent the majority of price-equivalent expenses
As AI agents handle increasingly complex tasks with extended context windows, the cost structure of LLM usage has shifted such that input tokens now represent the majority of price-equivalent expenses compared to output tokens. This trend reflects how modern AI applications are designed to process large amounts of contextual information upfront rather than generating longer outputs. The shift has implications for how developers should optimize their use of context-dependent AI tools and how pricing models are evaluated for cost efficiency.
Source: Cursor (X) | 2026-05-28