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 and image uploads, data analysis and voice each carry their own separate allowance, and storage is capped by plan. OpenAI publishes exact numbers for storage only; the message and upload allowances are deliberately undisclosed and change 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 then lists the tools that carry separate limits: data analysis, file and image uploads, and image creation. Voice and other tools are metered separately too. Hit one and that tool pauses until it resets; the chat keeps working. The Go article repeats the same structure — unlimited chat, higher tool ceilings than Free.
Now the split that most pages get wrong, in both directions. Message counts are not published at any tier — what OpenAI says about messages 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 notes it 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 cap at 20 files on Plus and 40 on Pro, Team, Education and Business; a custom GPT takes 10 files for its lifetime.
Storage is published twice, and the two figures do not obviously agree. The file-Library rollout 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. Both are OpenAI's own. A related gap it 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 available. OpenAI's Pro article states there is no setting to increase or bypass a model's usage allowance, and — under its own heading — that support does not reset usage limits. Reaching a limit is a designed state, not a fault.
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; on Go, at Free-tier levels.
- Individual models — on Pro, some models carry allowances that differ between the two Pro tiers.
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 release notes describe a smaller Instant Mini model that serves as "the fallback model users reach after hitting their rate limits" for the standard Instant model — and note that because it is a fallback, it does not appear in the model picker.
The practical consequence is worth knowing: a session that suddenly feels shallower may not be a bad day for the model you chose. It may be a different, smaller model answering because you spent the allowance on the first one. There is no banner for this.
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.
Two useful facts: files stay in your Library until you delete them, and files uploaded in a Temporary Chat are never saved to Library at all. If you routinely upload material you do not need to keep, Temporary Chat keeps the shelf clear as a side effect. Clearing space is Settings → Storage, or deleting files directly in Library — and deleting the conversation is what removes files and connected-app data together, as covered on our delete-history page.
In order of how much they cost you:
- Wait. ChatGPT shows the reset time in-product where one is available. Most limits reset on a timescale of hours.
- Switch tools. The chat itself is still unlimited — only the metered tool paused.
- Upgrade, remembering that upgrading restarts your billing cycle immediately. Doing that to skip a reset that arrives tomorrow is the expensive way to wait.
What does not work: contacting support. OpenAI states they do not reset ChatGPT or Codex usage limits, and that there is no setting to bypass one. If the limit persists past 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 2026-07-16. Note that only one of these actually removes the concept of an allowance.
Frequently asked.
Quick follow-ups people search after this question.