Billing and limits glossary
Terms for plans, prices, limits and spend. Shorter definitions of the core API terms are on the main Glossary.
Auto-reload
Buying credits automatically when the balance falls below a trigger. See billing.
Batch discount
The lower price for requests submitted as a batch and processed within 24 hours. See batches.
Bucket
One time slice in a report, such as an hour or a day.
Cost report
Dollar spend over time by workspace and line item; reconciles with invoices. See get.
Fast rate
The price for speed: "fast" output. See fast output.
Flat pay-as-you-go rate
The per-token price a plan applies to credit-paid usage of a model without its own price row.
Flex
Proposed — not offered: a lower-priority tier that may wait up to 10 minutes to start. No price is set. See service tiers.
Long-context rate
Proposed — not offered: a possible separate rate for requests whose input exceeds 262,144 tokens. No rate is set. See extended context.
Plan level
Base, middle or top. Each model has a minimum level. See models#plan requirements.
Priority commitment
Reserved tokens per minute on a model, billed monthly, served ahead of standard traffic. See priority capacity.
Rate-limit tier
A level of request and token-per-minute limits that grows with spend. See rate limits.
Region surcharge
Proposed — not offered: a possible premium for pinning processing to us or eu. None is set. See data residency.
Service tier
How a request is queued: standard, priority, flex or auto. See service tiers.
Spend limit request
A request to raise a workspace’s spend limit, approved by an admin. See create request.
Spill-over
Requests that asked for priority but ran on standard because the reserve was full.
Usage report
Token and request counts over time, grouped as you choose. See get.