TokenMaxxing: How to Get the Most Out of Your AI Models

In brief

• Anthropic's and OpenAI's top models cost 100 times more per token than their smallest ones, yet many users default to the flagship for everything.
• Usage leaderboards at Meta, Amazon, and Uber pushed AI bills up, and Uber burned through its 2026 AI tools budget in about four months.
• Routing tasks to the right model, trimming context, caching and batching cut costs sharply, but there are other techniques to save money.

Meta employees burned through 60 trillion AI tokens in 30 days earlier this year, chasing spots on an internal leaderboard nicknamed "Claudeonomics." Tech columnist Gergely Orosz estimated the tab at roughly $900 million at Anthropic's list prices, though Meta likely pays far less.

Bain & Company consultant Jue Wang sees a quieter version of that waste everywhere: "So many companies, so many users, default to using Opus for everything, including generating emails," she told the Associated Press this summer, referring to Anthropic's high-end models. If you do the same thing, you're hiring Albert Einstein to calculate your change at a MacDonald’s.

Every word you type into ChatGPT or Claude, and every word it writes back, is metered in tokens, chunks of text that average about three-quarters of a word. On a $20 subscription that meter decides how fast you hit your usage limit. On the API, the pay-per-token access developers use to plug models into their own apps, it decides the bill.