Model Pricing

AI model pricing questions: markup, per call, cache, free tier

Does BlockRun mark up model prices?

Not on tokens. Every chat model on this page is listed at the provider's own input and output rate per million tokens; the margin is a flat per-call transaction fee on the pay-per-call door, not a percentage on the tokens.

How is the price of a call quoted before it runs?

The first request comes back as an HTTP 402 carrying the exact dollar price for that call. The client signs that amount and resends; nothing is charged until the price has been seen.

Is there a free tier on the pricing page?

Yes. The models marked free need no wallet, no payment header and no signup — they are rate-limited per IP instead of billed. Every other row is paid per call.

What do images, video and speech cost?

Per image at the size requested, per second or per clip of video depending on the model, per thousand characters of speech, and per track of music — each shown on its own tab above, and each quoted in the 402 before the call runs.

Are cached tokens priced differently?

Where the vendor publishes a cache rate, the machine-readable sheet at /api/pricing carries it per model — a cache read rate and, where a separate write is charged, a cache write rate — beside the input and output rates, so a reseller or a ledger can bill cached tokens at the vendor's own number.

What about long-context pricing above a model's threshold?

Models whose vendor reprices above a prompt-size threshold carry that tier on the sheet: the threshold, whether it is inclusive, and the tier's input and output rates. The gateway bills the whole request at the tier the prompt reaches, exactly as the vendor does.

Can I pay with an API key and an invoice instead of per-call pricing?

Yes — pay per call with no account, or get an API key and a monthly invoice. The invoiced door meters exact usage at these same rates with no per-call minimum and bills in dollars monthly.

BlockRun pricing vs Together, Groq or the provider consoles — is it cheaper?

For tokens it is the same list price the provider charges, with no subscription and no prepaid balance on either door. Together and Groq price their own hosted models; the provider consoles price one vendor each. This page prices every vendor's models side by side, so the comparison is on the page rather than across tabs.

How do I read the price sheet programmatically?

GET /api/pricing returns every model with its rates per million tokens, its long-context tier and its cache rates as JSON; GET /v1/models returns the OpenAI-shaped list with a pricing block per model. Both are free to call.