Text Generation

Text across reasoning modes.

Fast, high-reasoning, structured extraction, and longform — one capability surface, one integration path. Durable jobs on eligible models use deferred processing at 50% of standard token rates. Live families: /capabilities/text-reasoning, /capabilities/text-structured, and /capabilities/translation.

Call capabilities, not model names

Choose fast generation, high reasoning, structured extraction, or longform through stable capability keys.

Use one job contract

Dialogue tools and data pipelines share submission, status, errors, and receipts.

Deferred processing on durable jobs

Eligible job_async text work runs at 50% of standard token rates. Invoke and stream stay on standard processing. See features.deferredProcessing on the capability descriptor.