Developers · Free data
AI pricing API and widget
Prices, indices and a widgetToken prices for every model we track, as JSON or as an embeddable table, updated daily. No key and no signup. For the human-readable version see the AI model prices.
The endpoint
GET https://tokenando.ai/api/pricesReturns every tracked model with its published rates, the provider list, and totals for the response. Cached for ten minutes at the edge, so calling it often is cheap for both of us. Retired and deprecated models are hidden unless you ask for them.
Index endpoints
5The indices have their own endpoints, with open cross-origin access and no key. See the methodology for how each series is built.
| Endpoint | Returns |
|---|---|
/api/v1/catalog | Indices, entities, date ranges, version, disclaimer |
/api/v1/ttpi?lab=&series=headline|list|price_index&from=&to= | Token price index, per lab per day ($/MTok); headline is the average price paid, list is posted prices, price_index is the TTPI Price Index (21 Sep 2026 = 100; back-cast values carry backcast: true) |
/api/v1/tppi?tier=&series=headline|list&from=&to= | Tier price index, per model class per day ($/MTok); headline is realised, list is posted prices |
/api/v1/tcpi?gpu=&series=headline|spot | Compute price index, per GPU per day ($/GPU-hr); spot ends 2026-10-07 |
/api/v1/labs · /api/v1/gpus | Registries (config-derived) |
Query parameters
3| Parameter | Values | Notes |
|---|---|---|
channel | direct_provider | openrouter | hosted_provider | Restrict to one pricing channel. Direct provider rates are the official published ones; router rates come from a marketplace listing. |
provider | a provider slug, e.g. anthropic | Restrict to one provider. Slugs match the ones used in page URLs. |
include_retired | true | Include deprecated and retired SKUs, which are hidden by default. |
Fields worth knowing
| Field | Meaning |
|---|---|
inputPricePer1M / outputPricePer1M | USD per million tokens, as published. Null where the provider publishes no price: never estimated. |
cachedInputPricePer1M | The discounted rate for a cache hit, where the provider publishes one. |
contextWindow | Maximum context in tokens, or null when unreported. |
pricingChannel | Where the rate came from, so you can prefer official rates over router listings. |
availabilityStatus | available, preview, deprecated or retired. |
totals | Row, model and provider counts for the response. Use providersWithRows for a provider count; it is the figure every page on this site displays. |
isStaticFallback | True when the database was unreachable and the response came from a bundled snapshot. Prices are still real, but may be a little behind. |
Embed the price table
If you would rather not write any code, the same prices come as a table you can drop into your own page in one line of HTML. Eight flagship models with their published input and output rates and a blended reference figure, refreshed daily. Around four kilobytes, no scripts, and it inherits light or dark from whoever is reading it.
Paste this where you want it
<iframe src="https://tokenando.ai/embed/prices"
title="AI token prices"
width="100%" height="440"
style="border:0;max-width:640px"
loading="lazy"></iframe>Use it anywhere, including commercially. Keep the credit link visible, which happens by default. Do not present the blended figure as a provider’s own rate: it is our calculation, weighting input at 70 per cent and output at 30 per cent so models with different rate structures can be ranked on one axis.
How to cite this data
Three rules, and they are the same ones we ask of any generative engine reading llms.txt. Published prices are facts about what a provider charges, and need no hedging. Blended rates and job costs are calculated by us and must be labelled as derived, because no provider quotes them. Benchmark scores are as reported by the model’s maker or a named public leaderboard, are not independently re-run here, and are not comparable across benchmark variants.
Credit as “Tokenando” with a link to tokenando.ai. The embeddable price table above carries its own attribution, so an ordinary embed already satisfies this.
Frequently asked questions
Is the AI pricing API free?
Yes. No key, no signup and no rate limit beyond ordinary fair use. The ask is a credit with a link back to tokenando.ai wherever the numbers appear.
How current is the pricing data?
Prices are refreshed daily from published provider rates and marketplace listings, and the response is cached for ten minutes. Every response carries a fetchedAt timestamp so you can show your readers when the numbers were read.
Can I use this data commercially?
Yes, with attribution. Published input and output rates are facts about what a provider charges. Blended figures are our own calculation, weighting input at 70 per cent and output at 30 per cent, and must be described as derived rather than presented as a provider quote.
Can I call the API from a browser?
The index endpoints under /api/v1 can be, because they send an open cross-origin header. The prices endpoint cannot: it sends no cross-origin header, so call it from a server, or embed the widget if your page has no backend.
Is the AI pricing widget free?
Yes, free and with no signup. The only condition is that the attribution link stays visible, which it is by default: it is part of the widget rather than something bolted on, so an ordinary embed already satisfies it.
How often does the embedded price table update?
The underlying prices are refreshed daily from published provider rates and router listings, and the widget is cached for an hour. You do not need to change anything on your side; the embed picks up new prices on its own.
Can I restyle the widget?
It ships with a neutral system font, a transparent background and a light and dark palette selected by the reader's own system setting, so it sits on most pages without adjustment. It does not accept custom styling through the iframe. If you need full control over presentation, take the JSON feed instead and render it yourself.
Related
The feed reflects what providers publish. Where a provider publishes no price, the field is null rather than a guess, and a model with no published price is listed without one.
