About OfficialCost

Cloud pricing is deliberately hard to compare. Providers bill on different axes, quote different units, bundle allowances that only apply to some plans, and publish rates across pages that assume you already know which product you need. Two services that look within a dollar of each other on their pricing pages can differ tenfold on an actual monthly invoice.

AI model pricing has the same problem in a newer form. Everything is quoted per million tokens — a unit nobody budgets in — with separate rates for input, output, and cached input, and the ranking between two models inverts depending on how much text they generate.

It holds 77 cloud services and 25AI models in one dataset with a consistent shape, applies each provider's own published rates to the same fixed workloads, and shows the working. That produces 499 head-to-head comparisons where the answer is a number rather than an opinion — and where the winner genuinely changes with volume, which is the part a rate card cannot tell you.

Every figure links to the vendor page it came from and carries the month it was checked. If the two disagree, the vendor is right.

The whole site is a static build. There is no database, no request-time API call, and no live scraping — the prices are checked into a set of JSON files, validated at build time, and rendered into flat pages. A malformed price fails the build rather than reaching a reader.

That is an unusual choice for a site with several hundred pages, and it is the right one: the data changes monthly, not by the second, and a static build removes every failure mode where the site is up but the numbers are not.