Strong AI text gets almost free
The price of getting a strong AI answer fell more than 280-fold in under two years, pushing the real value from answers to judgment.
A few years ago, asking powerful AI at scale felt expensive. Now the answer itself is cheap enough to spray into every app, inbox, search box, and workflow.
The cost of a strong AI answer fell from $20 to 7 cents per million text pieces.
- Smaller AI systems are doing work that used to need larger ones.
- Hardware and software are getting more efficient.
- Competition pushes companies to cut prices and bundle AI into normal tools.
When answers get cheap, your value moves up one level: asking better questions, spotting bad answers, and knowing what deserves action.
Behind the numbersOpen
Stanford’s AI Index tracks the cost of querying a model that reaches GPT-3.5-like performance on a broad knowledge test. It reported a drop from $20 per million tokens in November 2022 to $0.07 per million tokens by October 2024. A token is a small piece of text, often part of a word. The 2023 point uses the public GPT-3.5 Turbo input price of $0.002 per 1,000 tokens, equal to $2 per million tokens. Caveat: prices vary by provider, input versus output, speed, quality, and caching, so the direction matters more than one universal price.