Does ChatGPT have a limit?
Yes — but not where people expect. Everyday text chat is unlimited on every tier including Free. What is capped is everything around it: image generation, file uploads, data analysis and voice each carry a separate allowance, and storage is capped by plan. OpenAI publishes the upload and storage numbers; it never publishes a Chat message count, at any tier, and says limits shift with demand.
The honest answer includes what OpenAI refuses to tell you.
Start with what is documented. The Free Tier FAQ states that free users have unlimited everyday text chats, subject to abuse-prevention safeguards, and that file uploads, image generation, voice, data analysis and other tools have separate usage limits. Hit one and that tool pauses until it resets; the chat keeps working. The Go article repeats the structure — unlimited chat, higher tool ceilings than Free.
Now the split that most pages get wrong, in both directions. Chat message counts are not published at any tier — what OpenAI says is qualitative ("higher model limits", limits that "may vary based on system conditions", caps that bite "especially during high demand"), so any page quoting Plus as exactly N messages every M hours is stating a figure OpenAI does not. Upload limits, by contrast, are published in detail: up to 80 files every 3 hours, with Free users limited to 3 uploads per day, and OpenAI may lower these at peak. Per file there is a hard 512 MB ceiling, 2M tokens for text and documents, roughly 50 MB for spreadsheets and 20 MB per image; Projects take 5 files on Free, 25 on Go and Plus, and 40 on Pro and the organisation plans; a custom GPT takes 10 files for its lifetime.
Storage is published twice, and the two figures do not obviously agree. The Library article gives per-plan storage as Free 500 MB, Go 4 GB, Plus and Business 20 GB, Pro 100 GB. The File Uploads FAQ separately states a cap of 25 GB per end-user and 100 GB per organisation. We are not going to reconcile those by picking one — they sit in different articles describing different mechanisms, and the safe reading is that whichever is smaller for your plan is the one you meet first. A related gap OpenAI admits: ChatGPT shows Library storage in Settings → Storage but does not currently show how much of your rolling upload-rate quota remains.
Finally, the ceiling behaves the same way on the most expensive plan. OpenAI's Pro article states you cannot bypass a limit or increase it through a setting, and that Support cannot reset ChatGPT or Codex limits. The one exception is commercial, not technical: for Codex and ChatGPT Work, Plus and Pro can buy credits once the allowance is spent, or pay for an instant reset that pulls the weekly allowance forward. The Chat limits have no such option.
Which limit you actually hit.
The mental model that causes confusion is a single account-wide quota. There isn't one. There is a set of independent meters:
- Everyday text chat — unlimited on all consumer tiers.
- Image generation — its own allowance.
- File and image uploads — their own allowance.
- Advanced data analysis — its own allowance.
- Voice — its own allowance. OpenAI's Go article still puts Go at Free-tier voice limits; its pricing page now marks Go's voice as expanded.
- Thinking and Pro models — their own limits, which differ by plan and between the three Pro tiers.
- ChatGPT Work and Codex — one shared allowance, metered over five hours and a week on Plus; the Pro plans currently have no five-hour window there.
This is why "ChatGPT is unlimited" and "ChatGPT stopped letting me work" are both true. The meter you emptied is not the one you were watching.
Some limits do not announce themselves as limits. OpenAI's GPT-5.6 article says that if you reach a Thinking limit, ChatGPT may continue with another available model — and on Pro 200, reaching the weekly GPT-6 Pro limit switches you automatically to GPT-5.6 Thinking at Medium. Its release notes are blunter about the older chat models: a smaller Instant Mini model serves users who hit their Instant limit and, because it is a fallback, does not appear in the model picker.
The practical consequence: a session that suddenly feels shallower may not be a bad day for the model you chose. It may be a different model answering because you spent the allowance on the first one. ChatGPT shows a reset time where one is available, but the switch itself carries no banner.
Storage is the limit people breach slowly rather than suddenly. It covers uploads and generated files, so a habit of asking for spreadsheets and images fills it without any single moment that feels like heavy use.
Three useful facts. Files stay in your Library until you delete them manually. Files uploaded in a Temporary Chat are never saved to Library at all, so it keeps the shelf clear as a side effect. And — contrary to OpenAI's own File Uploads FAQ, which says files go when their chat does — the Library article states that deleting a chat does not delete the files it saved to Library. To free space, delete them in Library itself — Settings → Storage shows how much you are using; our delete-history page covers the chats themselves.
In order of how much they cost you:
- Wait. ChatGPT shows the reset time in-product where one is available.
- Switch tools or models. The chat itself is still unlimited — only the metered tool or model paused.
- Buy past it — only for Codex and ChatGPT Work. Plus and Pro can add credits once the allowance is spent, or buy an instant reset; the reset starts a new weekly period the moment you next use Work or Codex, and generally is not refunded.
- Upgrade. On Free, OpenAI says upgrading to Plus, Pro or Business resets your usage rates. How a mid-cycle upgrade is billed is not documented any more — read the checkout.
What does not work: contacting support. OpenAI states it does not reset ChatGPT or Codex usage limits, and that no setting bypasses one. If access has not returned after the stated reset time, that is a support case — but it is a bug report, not a request for more quota.
Ways to stop hitting it.
Three routes, cheapest first. Prices from our verified plan sheet, checked 2026-10-02. Note that only one of these actually removes the concept of an allowance.
Frequently asked.
Quick follow-ups people search after this question.