Last updated ยท September 1, 2026
Official figures are quoted as official, everything else is labelled, and every page carries the date it was last checked.
Reviewed: . Figures marked official come from OpenAI’s help center; everything else is labelled as observed or approximate.
Sources
OpenAI’s help center (GPT-5.6 in ChatGPT, Business models and limits, file uploads FAQ, deep research FAQ, Codex rate card, flexible pricing), the pricing page and OpenAI announcements. When a statement comes from these we say so and link it.
The limit messages and reset times ChatGPT shows. These are the most accurate signal for any individual account and we always point readers there first.
Where OpenAI confirms a limit exists but publishes no number (images, voice, consumer reasoning), we say “capped, unpublished” rather than inventing a figure.
We do not describe workarounds that break OpenAI’s usage policies, such as multiple accounts, and we do not present third-party guesses as facts.
Cadence
Each guide is re-read against the official sources. The visible date, structured-data date and sitemap date change together, only when content changed.
When OpenAI announces or ships a limit change, the changelog and affected guides are updated.
Found an error? Write to us with a link to the source. Corrections are applied on the page and noted in the changelog when a figure changes.
Translated homepages are summaries of the English guide and are updated in the same pass; the English pages are the reference.