What your AI spend actually costs, per million tokens.
Published list prices for 85 models across 14 providers and gateways — input, output, cached input, batch and context — each linked to the provider’s own page and dated. Model the bill before you sign it.
Rate table
All prices USD per 1M tokens unless stated. Sort any column. “Batch” shows batch/flex input where published.
| Source · checked |
|---|
Cheapest model that meets your context requirement
| # | Model | Context | Monthly cost |
|---|
Text models with a published context window only. Standard rates, no caching.
The weekly digest
Every Monday: rate changes, new models, promotions ending, and the cheapest option per tier — each with its source.
One email a week: what changed in AI API pricing, with sources. No spam; unsubscribe anytime.
Questions
Where do these prices come from?
Each row is read from the provider’s own published pricing page or official API (OpenRouter’s public models endpoint for OpenRouter rows). Every row links that source and shows the date we checked it.
What does “Unverified” mean?
We could not read a price from the official page on the check date (for example, it is rendered client-side, region-dependent, or listed as “Contact sales”). We show the row so the gap is visible, and never guess a number.
How is the calculator’s cost computed?
Monthly cost = uncached input × input rate + cached input × cached rate + output × output rate, blended with batch rates for the share of traffic you mark as batch. It excludes tool calls, cache-write premiums, long-context surcharges, taxes and negotiated discounts.
How often is the data checked?
A script snapshots every source page weekly and diffs it. Confirmed changes are written to the change log and the RSS feed.
Can I use the data in my own tools?
Yes. The JSON API, OpenAPI spec and llms.txt are free to use with attribution.