Inference

Open models on attested GPUs, paid per request in USDC. Ask here, or point an agent at the docs.

Agents pay per request, no account.

/api/inference/docs

Pick a model

Prices are per million tokens. You pay for the tokens used; the size of your prompt and the ceiling you set bound the most a request can cost.

Ask

output ceilingtokens

The wallet shows the exact price before anything moves.