A small document-processing tool stops accepting requests, even though its account has credit. Before buying more, inspect which limit it reached. A burst of requests and a monthly spending cap require different responses.
What changed on 6 October
OpenAI reduced its paid API usage tiers from five to three: Build, Launch and Grow. They belong to API organisations, not ChatGPT subscription names. Read the tier shown for the organisation actually running your tool. This is an explanation of the 6 October change, checked on 14 October, not a new announcement today.
Check three different numbers
- Rate limits describe requests or tokens over time for the selected model.
- Monthly usage limits describe the organisation’s allowed usage.
- Your own spending controls can set a separate ceiling or notification.
In the API dashboard, open the organisation’s Limits settings and check the selected model. A spending alert sends a warning; a hard spending limit can stop affected requests. Do not assume a notification also blocks spending.
Try a smaller, observable run
For a fictional batch of twelve documents, first process one and record the outcome. If the error is a temporary rate limit, reduce simultaneous work and respect the reported retry delay. If it is a spending limit, inspect that control instead of repeatedly resending the request.
Keep a record of completed document IDs so a retry does not create duplicate outputs. Resume only the unfinished items, then inspect one saved result. Upgrading throughput does not prove a lower bill or better answer quality.