.png&w=1536&q=75)
Free Lesson
Where the tokens go: the cost of an LLM API call
120 min
Dec 19, 2026 12:00 PM
By continuing, you agree to Maven's Terms and Privacy Policy.
What you'll learn
Where the tokens come from
Some of the tokens you pay for are ones you never wrote, and output and reasoning tokens cost more than input.
How caching shows up in your bill
Providers charge less for the start of a prompt they’ve seen recently, if it stays the same between requests.
Which changes bring the cost down
Reorder the prompt so caching works, send less text, ask for shorter output, pass easy requests to a smaller model.
Why this topic matters
LLM costs tend to rise as a product grows, as prompts get longer and teams pick a bigger model because it’s the easy choice. A bill that looks fine in testing can become a problem at volume. Knowing where the tokens go lets you cut costs with measured changes, and check that the cheaper version still does the job.
You’ll get the app’s repository before the session, so you can follow along or run each step yourself afterwards.
.png&w=384&q=75)
