Skip to main content
The GPU Pricing API serves Compute Index market data through one consistent surface. Every data type answers the same four questions the same way, using the same filter names and returning the same response shape. This is v2, which supersedes the v1 API — new integrations should start here. Base URL: https://data-api.compute-index.com/v2
Already using v1? Every v1 endpoint has a v2 replacement, and the same token calls both — see Migrating from v1.

​
The four endpoints

Every data type exposes the same four, or explicitly declares that a question does not apply to it. Learn one data type and you have learned them all — see Concepts for how they behave.

​
Data types

GPU Rental Prices

What it costs to rent GPU compute, across 50+ vendors, in USD per GPU-hour.

GPU Rental Indexes

Blended GPU price indices, daily and at 30-minute resolution, by GPU family or by GPU. One index per request; /metadata lists the ones you hold.

Aggregated Transactions

Settled deals combined in groups, so no single deal can be identified.

Hardware Prices

What it costs to buy GPU hardware outright. New in v2.

Token Prices

LLM inference prices per model, in USD per million tokens. New in v2.

​
Indices

Every published index is served from one place, and a request names one index by id. /metadata lists the ids your token holds, each with its cadence, its row shape and who computes it — see Indexes for the catalogue and the two row shapes.

​
Coming from v1?

Every v1 endpoint has a v2 replacement, and the two can be called side by side with the same token while you move across.

Migrating from v1

Endpoint-by-endpoint mapping, how to swap your API key for an access token, and the four behaviour changes to check.