Since August 2026 OpenAI has listed everyday text chat as unlimited on every ChatGPT plan, subject to abuse guardrails. The limits moved rather than went away. The Medium and High thinking levels, file uploads, deep research, and ChatGPT Work and Codex all have allowances of their own, and OpenAI only publishes numbers for some of them. When someone says they've hit their ChatGPT limit, it's almost always one of those. The habits that stretch them are the same four that stretch any AI plan.
What's actually capped in ChatGPT?
Ordinary chat on the Instant setting isn't. Almost everything you'd use for serious work is, each on its own clock:
What's capped in ChatGPT
| What | Limit | Resets | Worth knowing |
|---|---|---|---|
| Everyday chat (Instant) | Unlimited | n/a | Subject to abuse guardrails |
| Thinking: Medium and High | Depends on plan; number not published | ChatGPT shows when, if it can | Plus doesn't get Extra High or Pro |
| File uploads | 80 every 3 hours; 3 a day on Free | Rolling | Can drop at peak times. Failed uploads can count. No meter |
| ChatGPT Work and Codex | About 15–150 messages per 5 hours on GPT-6 Sol, 5–45 on GPT-6 Astra | 5 hours | One shared allowance, separate from chat |
| Pro model (Business) | 15 a month on Standard; 50 a week on Premium | Monthly or weekly | Shared across GPT-6 Pro and GPT-5.6 Sol Pro |
The Thinking limit is the one most people run into, and the one with no number attached. OpenAI says only that it depends on your plan. When you reach it, ChatGPT may carry on with another model and tells you when your allowance resets, if it can. Since 14 September Plus and Pro no longer switch into Thinking on their own, and typing "think harder" doesn't change the level either. You pick it in the model picker.
Look at the Work and Codex range too. Fifteen to 150 messages in five hours is a tenfold spread on the same seat. The difference is how much each task asks for, which is exactly what the habits below control.
If chat is unlimited, why do long chats still matter?
Because the meter isn't the only cost. Every reply means ChatGPT reading the thread above it. On Instant that doesn't come off an allowance, but it still catches up with you in three ways. Very long threads eventually hit a maximum length and ChatGPT asks you to start a new chat. Answers drift as early instructions get buried under later ones. And in ChatGPT Work and Codex, OpenAI says usage depends on the size and length of the session, so a sprawling session costs more of the five-hour allowance.
A new chat doesn't reset any allowance, though. It only resets what ChatGPT has to read.
Where does your ChatGPT allowance go?
We sort it into four things. We call them Intersect's Four Ls, and they apply to every assistant we use.
Intersect's Four Ls
How long the thread has run.
Long threads hit a maximum length, drift off the brief, and make Work and Codex tasks cost more.
One task per chat. Move on with a handover note.
Everything you attach or connect.
Uploads are capped at 80 every 3 hours, and re-attaching the same file spends another one.
Put repeat files in a Project once. Send the section, not the deck.
The thinking level you pick.
Instant is unlimited; Medium and High aren't. Since September you choose the level yourself.
Instant by default. Step up per question, not per chat.
The rounds it takes to get it right.
Every retry on High spends a capped message and re-reads the thread.
Brief properly. Edit the prompt instead of correcting it.
How do you cut the length?
One task, one chat. When the task changes, start again. To carry work across, ask for a handover note, check it, and paste it into the new chat:
We're moving to a new chat. Write a handover note I can paste into it. Include: the goal, what we've decided and why, the latest version of the work (or where it lives), anything still open, and the next step. Leave out anything we tried and dropped. Keep it under 300 words.
ChatGPT's memory and its ability to reference past chats in the same Project help here too. You can point it at earlier work without pasting it back in.
How do you lighten the load?
- Use Projects for repeat files. Upload your price list, brand guide or template to a Project once, and every chat in it can use them. Re-attaching the same file to each new chat spends another upload every time.
- Send the part, not the pile. Ask about slide 12, not the 60-slide deck. Ask about one tab, not the workbook.
- Know the upload clock. It's 80 files every three hours on paid plans and three a day on Free. OpenAI may lower it at busy times, failed uploads can count against it, and there's no meter showing what's left.
How do you match the lift to the job?
Instant is the unlimited part of ChatGPT, so make it the default and step up only when a question needs it. Here's how we'd set it for common jobs:
Match the lift to the job
| Job | Thinking level | Tool |
|---|---|---|
| Rewrite an email, summarise a page, quick answer | Instant | None |
| Draft a proposal or report from your notes | Instant, or Medium for structure | None |
| Analyse a spreadsheet, contract or model | High | File upload |
| Check a current fact | Instant | Search |
| A sourced report across many sites | n/a | Deep research |
| A multi-step task across your apps | n/a | ChatGPT Work (uses the Work and Codex allowance) |
Change the level per question, not per chat. A thread that switched to High for one hard question and stayed there spends a capped message on every "thanks, now make it shorter" that follows.
How do you cut the loops?
- Brief it properly. Audience, length, purpose, what good looks like and what to avoid, in one message.
- Batch related questions. One message with five questions beats five messages.
- Edit, don't correct. Edit your original prompt and ChatGPT answers the new version on a new branch, without the wrong answer sitting in the thread.
Where does this break?
You can't plan around a number OpenAI doesn't publish. On Plus you find the Thinking limit by hitting it, and the only warning is the message telling you when it resets. If your team leans on High all day, Business or Pro is the honest answer, not better habits.
Instant isn't the right tool for everything. A hard analysis on Instant that takes four rounds of correction costs more time than one pass on High. Lift is cheaper than loops.
And "unlimited" has a footnote. OpenAI applies abuse guardrails to it, and its Plus page has long said limits can tighten when demand is high.
Six changes, ten minutes
- Set the thinking level back to Instant and step up only when a question needs it.
- Start a fresh chat for your next task. Bring anything you need over with the handover prompt.
- Move the files you attach every week into a Project.
- Send the slide, tab or clause you're asking about, not the whole file.
- Keep deep research for questions that need a sourced report.
- Next time ChatGPT misreads you, edit your message instead of correcting it.
Want your team getting more from the same plan?
We help teams set up ChatGPT, Claude or Gemini properly: Projects with your files and house rules, the right settings for each kind of job, and an hour with your people on habits like these. If you're hitting limits every week, we'll tell you whether it's the habits or the plan.
FAQs
Does ChatGPT Plus have a message limit?
Not for everyday chat on Instant. The Medium and High thinking levels, file uploads, deep research, and ChatGPT Work and Codex have separate limits.
Does starting a new chat reset my limit?
No. It resets what ChatGPT has to read, not your Thinking, upload or Codex allowance.
What happens when I hit the Thinking limit?
ChatGPT may carry on with another model and shows when your allowance resets, if that information is available. Instant chat keeps working.
Do Work and Codex share the chat limits?
No. They share one allowance between them, counted separately from chat, and on Plus it runs on a five-hour window.
Usage limits, assistant by assistant
Related reading
- Hitting Your Claude Limit? Here's Where the Tokens Go
- How to Prompt ChatGPT and Claude: An AI 101 for Your Team
- Set Up a Shared AI Workspace for Your Team in an Afternoon
Sources
- OpenAI Help Center: GPT-6 and other models in ChatGPT, accessed 6 November 2026
- OpenAI Help Center: ChatGPT Business models and limits, accessed 6 November 2026
- OpenAI Help Center: File Uploads FAQ, accessed 6 November 2026
- TechCrunch: OpenAI brings unlimited ChatGPT text chats to free users, 6 August 2026
- Prompt Optimizer: the ChatGPT thinking setting, 5 October 2026
- LaoZhang AI: ChatGPT usage limits by plan, updated 8 October 2026