What a credit actually costs across 50+ AI products — bundled rates, top-up premiums, and the expiry and exhaustion rules behind the meter, archived per row.
AI software is quietly migrating from seats to meters. Buy a plan and you
no longer get features — you get a monthly allotment of credits,
tokens, actions, or outcomes that deplete as you
work. The AI Credit Index is a quarterly study of how that meter is priced
and governed across 51 products, read
only from vendors’ own public pages and archived row by row. This is
edition 1.
The headline number is a warning, not an answer: the median bundled rate is
about 1.6¢ per credit, but published rates run
from $0.0000018 to $25
— a spread of more than 13 million
to one. A “credit” is not a unit; one buys a tweet in one product
and a feature-film shot in another. So the Index never compares raw
$/credit across products — it compares how products treat the same
credit.
What a credit costs, by category
category median (shown at n ≥ 5 products)
Each dot is one plan's bundled $/credit on a log scale (115 rates); aubergine marks each category's median, shown only where the category has ≥5 products. The shape is a map of unit-size: the cheap end meters tokens (coding, incumbents), the dear end prices whole outcomes or seats (support/CX, dev platforms).
SoF · The AI Credit Index
51products reviewed
196plan tiers analyzed
+12%median top-up premium
39%of top-ups hide the price
Read this three ways
The Index is built in three layers. Take the facts as a
citable source; use the analysis to see the patterns; then
turn it into your own pricing decisions.
Or jump to Packaging (free tiers, grants,
annual terms), The fine print (expiry,
overage, what happens at zero), Transparency,
or What’s changing. For a single product’s
history — its rung, its moves since January 2025, and its likely next step —
see the per-product Reviewed pages. The full
panel, source ledger, and method live under
Methodology & data.
Related questions
What does a credit actually cost across AI products?
Across 51 AI products with public credit pricing, the median bundled rate works out to about 1.6¢ per credit ($0.01571). But that median is close to meaningless on its own: published bundled rates span from $0.0000018 to $25 per credit — a spread of more than 13 million to one — because a "credit" buys a wildly different amount of work in each product. The Index therefore compares how products treat the same credit, not the raw price of one.
Is it cheaper to buy a bigger plan or to top up?
Usually the plan. Of the 24 plans where both the bundled rate and the top-up price are public, 13 charge more per credit to top up than the plan's own bundled rate, and the median top-up premium is +12%. Premiums run as high as +350% (the entry tier of one design tool); only 7 make topping up cheaper than the bundle, and 4 price it identically — so topping up never actually beats the plan.
Do AI products disclose what extra credits cost?
Often not. Of the 133 plan tiers that offer a top-up, only 61% publish the price on a pricing page or in help docs; for the rest the price appears only at the in-app purchase screen, after you've subscribed. A further 13 category-leading vertical-AI products publish no pricing page at all.
How is the AI Credit Index sourced and verified?
Every figure comes from a vendor's own public pricing page or help docs — never third-party trackers, leaks, or estimates. Each published product is captured to a dated observation backed by at least one Wayback Machine snapshot, and all derived figures are computed from those observations, never hand-entered. Where a price is hidden behind the logged-in purchase screen, that is recorded as hidden rather than guessed: the disclosure rate is itself one of the things being measured.
