DeepSeek overtakes Google on volume, cost per token falls 13.6%
Source
vercel.com
Author
Amelia Charles
Date
Context
The average cost of AI text generation can fall while total spending rises if customers process more text and choose cheaper models. Vercel's August 2026 Production Index reports exactly that pattern in its AI Gateway traffic for July: tokenThe chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.Full definition → volume grew 59%, calculated spending grew 37% and average cost per token fell 13.6%. Holding the previous month's model mix constant would have left average prices roughly unchanged.
The report's methodology values requests at published list prices. Its result describes a shift in model selection within this gateway, rather than a measurement of every customer's negotiated bill or the whole AI market.
Terms in this piece · Glossary
token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.
