About What Is a Million Tokens
AI pricing is confusing. "Per million tokens" means nothing to most people. We translate token counts into real-world terms everyone understands — novels, pages, lines of code, and dollars.
Our Mission
What Is a Million Tokens exists to help developers and businesses understand AI token pricing in plain language. Whether you're evaluating GPT-5, Claude, or Gemini for your next project, you shouldn't need a spreadsheet to figure out what it'll cost.
We compare every major AI model side by side — input and output pricing, context windows, and real-world equivalents — so you can make informed decisions in minutes, not hours.
All data on this site is fact-checked against official provider documentation. We link to primary sources whenever possible so you can verify everything yourself.
How We Keep Prices Current
AI pricing changes fast. Models get cheaper, new tiers launch, and promotional rates come and go. Our pricing data is updated daily via automated checks against the OpenAI, Anthropic, and Google APIs and pricing pages.
When a provider updates their pricing, our system detects the change and flags it for review. We verify every update before it goes live on the site.
How We Calculate
Every estimate on this site uses the standard rule of thumb:1 token is roughly 0.75 English words (or about 4 characters). This ratio has been verified using actual tokenizers from OpenAI (tiktoken), Anthropic, and Google across thousands of English text samples.
We always show our assumptions. If a comparison card says "10 novels," it also tells you we're assuming 80,000 words per novel. No hidden math, no hand-waving.
Keep in mind that token counts vary by language, content type, and model. Code and non-English text typically use more tokens per word. Our figures are calibrated for typical English prose and clearly labeled when a different assumption applies.
Who Built This
What Is a Million Tokens is an independent project by an AI developer. It is not affiliated with OpenAI, Anthropic, Google, or any other AI provider. We don't receive compensation from any model provider, and our comparisons are unbiased.
The site is open source. You can view the code, suggest corrections, or contribute on GitHub: