The measure of how effectively an AI model utilizes input and output tokens to complete a task, where higher efficiency can reduce total costs even if per-token prices are higher.
Overview
- [Industry analysis] Switching to cheaper AI models, particularly Chinese models, does not necessarily deliver bottom-line savings because more expensive models may be more efficient in token usage, resulting in lower total costs for specific tasks. (Source: Alpha Sense CEO Jack Kokko cited by AI Daily Brief host, via AI Daily Brief, 2026-08-24)