Token Price Index

The Token Price Index is a number that shows, at a glance, how much the fees you pay to use AI models have risen or fallen over time — much like a consumer price index. Just as a CPI is built by pooling the prices of many goods at a store, this index is calculated by pooling the token (text-chunk) fees charged by various AI companies to capture the average trend.

There's been a lot of talk lately about this index dropping sharply over the past few months, driven by companies like OpenAI and Anthropic competitively cutting their model prices. As more businesses switch to cheaper models coming out of China, established companies are facing pressure to lower their prices to keep up.

This index matters because for services that call an AI model dozens of times rather than just once, even a small drop in token fees can have a big impact on overall operating costs. This is especially noticeable in services like coding assistants, which invoke a model many times to complete a single task.

The article mentions that "the Token Price Index tracked by Silicon Data has fallen about 25% since mid-July." This refers not to the price of any single company, but to an average decline pooled across multiple models' fees. A common point of confusion is mistaking this index for a performance metric — but it's actually a number that purely reflects fee (cost) trends.