Skip to content
← Writing
Writing · October 2026 · 2 min read

Cheap tokens, expensive bill

On Vercel's AI Gateway, the price per token fell 47% while spend rose 13 times. Price the task, not the token, and keep your model swappable.

Abhijit Chatterjee Chattopadhyay

The Logic Atelier logo above the words: Cheap tokens, expensive bill. Price the task, not the token.

AI got cheaper. The bills got bigger.

On Vercel's AI Gateway, the average price per token fell 47% between December and September. Spend through the gateway rose about 13 times over the same months. That is from Vercel's AI Gateway Production Index for October 2026 (opens in a new tab), which tracks real production traffic.

Cheap tokens do not make a cheap system.

Where the money really goes

The same report shows the split clearly. Open-weight models now handle more than seven in ten tokens on the gateway, but take nearly one in four dollars.

One model, DeepSeek-v4.1-Flash, processed nearly half of the gateway's September tokens for just 2.6% of spend. Meanwhile, the more expensive models drove most of the growth in spend.

So the cheap model is doing the volume. The expensive model is doing the work that costs.

Teams already mix models

Serious teams do not pick one model and stop. On the gateway, teams that processed at least 10 million tokens since December used 15 models on average. Teams past a billion tokens used 42.

And they switch. In September, nearly one in two teams changed their most-used model.

The model is becoming a replaceable part. Your architecture should treat it that way.

Price the task, not the token

For each AI task in your product, ask three questions:

  1. What does a wrong answer cost? A typo in a summary is cheap. A wrong figure in a quote is not. High cost of error earns a stronger model.
  2. How often does it run? A task that runs a million times a day earns a cheap, fast model and careful testing.
  3. Can you check the output? If a rule or a test can catch a bad answer, a cheaper model plus the check often beats an expensive model alone.

Then route each task to the cheapest model that passes your tests for it. Not the cheapest model on the price list.

Do this today

  • List your AI tasks and the model each one uses today.
  • Write down the cost of a wrong answer for each, in plain money terms.
  • Put the model name in configuration, not in code, so switching is a setting, not a project.
  • Re-test your top three tasks on one cheaper and one stronger model. Keep the results.

Your next step

If you want your AI costs under control, book a free 30-minute 1:1 assessment. I look at your setup with you and tell you what the work would take.

Drafted with AI help, edited and approved by me.

New articles by email

I will never sell your details, for any reason, and I never send spam.

How I use your details
  • I, Abhijit Chatterjee Chattopadhyay of The Logic Atelier, hold them, in my own newsletter system hosted on Fly.io. Emails go out through Resend.
  • Your name is for greeting you, your email is for sending new articles, and your phone number, if you give it, is so I can reach you directly about my work.
  • If you do not confirm your email within 7 days, I delete the sign-up. If you unsubscribe, I delete your details 30 days later.
  • If Google Analytics is running on your visit, it is told that a sign-up was made, and nothing you typed.
  • To see or delete what I hold about you at any time, email hello@thelogicatelier.com.
  • The full detail is on the privacy page.
NextYour AI agent can now pay. Who set the budget?
← cd ../writing

Cheap tokens, expensive bill

cheap-tokens-expensive-bill.mdfront matter
title: Cheap tokens, expensive bill
date: 2026-10-18
reading_time: 2 min
author: Abhijit Chatterjee Chattopadhyay
cheap_tokens_expensive_bill_cover_e1d3d705_2c7d03a2e6.pngcover
The Logic Atelier logo above the words: Cheap tokens, expensive bill. Price the task, not the token.

AI got cheaper. The bills got bigger.

On Vercel's AI Gateway, the average price per token fell 47% between December and September. Spend through the gateway rose about 13 times over the same months. That is from Vercel's AI Gateway Production Index for October 2026 (opens in a new tab), which tracks real production traffic.

Cheap tokens do not make a cheap system.

Where the money really goes

The same report shows the split clearly. Open-weight models now handle more than seven in ten tokens on the gateway, but take nearly one in four dollars.

One model, DeepSeek-v4.1-Flash, processed nearly half of the gateway's September tokens for just 2.6% of spend. Meanwhile, the more expensive models drove most of the growth in spend.

So the cheap model is doing the volume. The expensive model is doing the work that costs.

Teams already mix models

Serious teams do not pick one model and stop. On the gateway, teams that processed at least 10 million tokens since December used 15 models on average. Teams past a billion tokens used 42.

And they switch. In September, nearly one in two teams changed their most-used model.

The model is becoming a replaceable part. Your architecture should treat it that way.

Price the task, not the token

For each AI task in your product, ask three questions:

  1. What does a wrong answer cost? A typo in a summary is cheap. A wrong figure in a quote is not. High cost of error earns a stronger model.
  2. How often does it run? A task that runs a million times a day earns a cheap, fast model and careful testing.
  3. Can you check the output? If a rule or a test can catch a bad answer, a cheaper model plus the check often beats an expensive model alone.

Then route each task to the cheapest model that passes your tests for it. Not the cheapest model on the price list.

Do this today

  • List your AI tasks and the model each one uses today.
  • Write down the cost of a wrong answer for each, in plain money terms.
  • Put the model name in configuration, not in code, so switching is a setting, not a project.
  • Re-test your top three tasks on one cheaper and one stronger model. Keep the results.

Your next step

If you want your AI costs under control, book a free 30-minute 1:1 assessment. I look at your setup with you and tell you what the work would take.

Drafted with AI help, edited and approved by me.

subscribeemail

// I will never sell your details, for any reason, and I never send spam.

How I use your details
  • I, Abhijit Chatterjee Chattopadhyay of The Logic Atelier, hold them, in my own newsletter system hosted on Fly.io. Emails go out through Resend.
  • Your name is for greeting you, your email is for sending new articles, and your phone number, if you give it, is so I can reach you directly about my work.
  • If you do not confirm your email within 7 days, I delete the sign-up. If you unsubscribe, I delete your details 30 days later.
  • If Google Analytics is running on your visit, it is told that a sign-up was made, and nothing you typed.
  • To see or delete what I hold about you at any time, email hello@thelogicatelier.com.
  • The full detail is on the privacy page.

vi.Contact

06 · contact

Talk to me.

Talk to me.

Production AI and software architecture. Based in Dubai, working with teams in the Emirates and Ireland.

I work Dubai hours, Monday to Friday, the same week you do. Dubai is three hours ahead of Dublin in summer and four in winter, so my afternoon is your morning.

I will never sell your details, for any reason, and I never send spam.

How I use your enquiry
  • Your enquiry comes to me, Abhijit Chatterjee Chattopadhyay of The Logic Atelier. The form is handled by Formspree, which passes it to my inbox and keeps a copy in my Formspree account.
  • I use it to read your enquiry and reply. If no work follows, I delete it within 12 months, and sooner whenever you ask.
  • If Google Analytics is running on your visit, it is told that an enquiry was sent, and nothing you typed.
  • To see or delete what I hold about you, email hello@thelogicatelier.com.
  • The full detail is on the privacy page.