How-to

Which ChatGPT Plus limits Pro actually fixes

ChatGPT limits Chat separately from Work and Codex, and Pro lifts some limits more than others. Paste a week of limit messages into this prompt to see whether Plus, a month of Pro or a cheaper fix suits you.

Pro now comes in three monthly tiers, US$100, US$200 and a new US$500, against US$20 for Plus, though prices vary by country. The biggest gap is in Work and Codex, where OpenAI says Pro plans currently have no five-hour limit, so the limits you hit in a normal week decide which plan pays.

The prompt: read your week, get a verdict

Run it in any AI chat. Screenshot or copy each limit message word for word when it appears, because the wording decides which limit it was.

Prompt Decide between Plus and Pro
You are helping me decide whether to move from ChatGPT Plus to ChatGPT Pro, using only what I actually hit in a normal week. Work like a careful friend who has used both plans and has no reason to upsell me. If you push me to Pro when a cheaper fix would do, I pay at least US$80 a month more for nothing. If you keep me on Plus when my limits are costing me real work, I lose hours every week. So judge only my log, and say when it is too thin to judge.

FACTS TO USE (from OpenAI's own pricing and help pages, read on 12 November 2026; do not add to them from memory, and if today is much later, tell me to recheck them before I pay)
- Prices in US dollars a month, though prices vary by country: Plus US$20. Pro comes in three tiers: Pro 100 at US$100, Pro 200 at US$200, Pro 500 at US$500. Pro 200 includes more usage than Pro 100, and Pro 500 the most. No other Pro tier appears on OpenAI's pages. Pro is billed monthly only.
- ChatGPT has separate limits. Chat has its own limits by model and thinking level. Work and Codex share one allowance with each other, and Word, Excel and PowerPoint draw on it too. Images, deep research, voice and uploads have their own limits.
- Chat thinking levels: Plus includes Medium and High. Extra High and Pro thinking (GPT-6 Pro, powered by GPT-6 Astra) are not included on Plus and are included on Pro. The pricing page lists GPT-6 Sol, which runs Instant to Extra High, as Expanded on Pro, with no numbers for either plan.
- GPT-6 Pro has weekly limits on every Pro tier; a Pro subscription does not include unlimited use of it. On Pro 100, the two Pro models, GPT-6 Pro and GPT-5.6 Sol Pro, share one weekly allowance. On Pro 200, reaching the GPT-6 Pro weekly limit switches ChatGPT to GPT-5.6 Thinking at Medium.
- Work and Codex on Plus: OpenAI estimates 5 to 45 local messages per five hours on GPT-6 Astra and 15 to 160 on GPT-6.1 Sol, and says these are estimates, not fixed limits. Weekly limits may also apply. Plus includes only limited Astra usage.
- Work and Codex on Pro: OpenAI says Pro plans currently have no five-hour limit. Weekly limits may still apply.
- OpenAI's business pricing page says that starting in December 2026, plan usage will share one limit across all of ChatGPT and Codex. If my log is from December or later, tell me the Chat and Work limits may now be one shared limit, and sort hits by what I was doing.
- Astra may use the allowance faster than other models. Fast mode uses the allowance at 2.5 times the normal rate. Ultrafast, on Pro 500 only, uses it at 8 times.
- OpenAI says a larger allowance does not guarantee twice as many completed tasks.
- Plus and Pro users who reach a Work or Codex limit can buy credits to keep going without upgrading, if credit purchases are available on their account (Settings, then Usage). Credits cover Codex, Work, Word, Excel and PowerPoint, not every model or feature, and last 12 months. Once bought, they are used automatically after the included allowance runs out; a saved instruction does not add an approval step. Credit prices are not on these pages.
- Context: Instant handles about 40 pages of text on Plus and about 250 on Pro. Reasoning handles about 320 pages on Plus and about 680 on Pro.
- Pro only, among personal plans: Dots (always-on agents), ChatGPT Space, Ultrafast (Pro 500), and monthly API credits of US$100, US$200 or US$500 by tier, which expire at the end of each billing cycle and have eligibility rules.
- Business costs US$25 per user a month billed monthly, needs at least 2 users, and its standard seats get the same Work and Codex estimates as Plus. Business Premium seats cost US$125 per user a month billed monthly and have no five-hour limit.
- Cancelling Pro: a scheduled cancellation keeps access to the end of the billing period and can be undone before then. Cancel where I bought it: ChatGPT, Apple or Google Play. Cancelling does not refund earlier charges.
- Hitting a limit does not restrict an account, and OpenAI Support cannot reset limits.

ABOUT ME (I fill this in once; any line can say "not sure")
- Where I live and pay: [country]
- Prices my checkout shows: [the Plus price and the Pro 100 price, e.g. "US$20 and US$100", or "not checked"]
- My plan now: [e.g. "Plus since March"]
- What I use ChatGPT for, most to least: [e.g. "Work for client reports, Chat for research, Codex a little"]
- Who the work is for: [e.g. "my own business", "my employer", "personal projects"]
- The most I will spend on ChatGPT a month: [e.g. "US$100", or "not sure"]
- What an hour of my working time is worth to me: [e.g. "US$40", or "skip"]
- Other AI subscriptions I already pay for: [e.g. "Claude Pro", or "none"]
- Whether my employer offers ChatGPT Business or Enterprise: [yes / no / not sure]

MY WEEK OF LIMITS (one line per hit, for at least 3 working days)
- [day and time] | [Chat / Work / Codex / Word, Excel or PowerPoint / images / deep research / voice / other] | [model and thinking or speed setting, e.g. "GPT-6 High", "Astra in Work, Fast on"] | [the exact message ChatGPT showed, including any reset time] | [what I was doing and what it cost me, e.g. "waited 3 hours, report went out late"]
- My usage page readings: [paste the percentages or reset times from Settings, then Usage, or write "not checked"]

HOW TO WORK
Check that ABOUT ME and MY WEEK are filled in. If MY WEEK is blank, ask for it once and wait. If it covers fewer than 3 days, say the verdict is low confidence and give it anyway, marked LOW EVIDENCE. Then do Jobs 1 to 5 in order.

Job 1. Sort every hit into one group, quoting the message that decides it.
- Group A, Chat thinking: a limit on a thinking level or a Pro model in Chat, or I wanted Extra High or Pro thinking and Plus does not have it.
- Group B, Work and Codex: a five-hour or weekly limit in Work, Codex, Word, Excel or PowerPoint.
- Group C, too long: a file or chat too big to read in full.
- Group D, tool limits: images, deep research, voice or uploads.
- Group E, not a plan limit: an outage, an error, or wording about my account being restricted. Tell me a plan change will not fix these.
- Group F, cannot tell: mark NOT STATED and ask me for the exact wording. Do not guess.

Job 2. For each group, say what Pro changes, using only FACTS TO USE, and the cheaper fix to try first.
- Group A: Pro adds Extra High and Pro thinking. If I only hit limits on Medium or High, try a lower level for routine questions first. Pro thinking still has weekly limits.
- Group B: Pro removes the five-hour limit. If Fast mode was on, the first fix is turning it off, since it uses the allowance at 2.5 times the rate. If the hits were on Astra, try GPT-6.1 Sol for routine tasks. If hits fall on 2 days a week or fewer, credits may cost less than an upgrade; tell me to buy a small amount once and note what it cost and how many hit-days it covered, then work out credit cost divided by hit-days covered, times hit-days a week, times 4.3, and compare that monthly figure with Pro 100's extra cost from Job 3.
- Group C: Pro raises the page limits above. If my files fit the Plus limits after splitting them, splitting is the cheaper fix.
- Group D: the pricing page lists unlimited and faster image creation and maximum deep research on Pro, and gives no numbers for Plus. Treat this as the weakest reason to upgrade unless these hits block paid work.
If my hits span several groups, rank them by what they cost me, from my own descriptions.

Job 3. Do the arithmetic, and show every step.
- Monthly extra cost over Plus: Pro 100 is US$100 minus US$20, which is US$80. Pro 200 is US$180. Pro 500 is US$480. If the checkout prices I gave differ from these, use mine and say so.
- If I gave an hourly value and hours lost: count only hours from hits the Job 2 cheaper fixes would not remove, and never from Group E. Hours lost a month is those hours from my week times 4.3. Break-even hours for each tier is its extra cost divided by my hourly value. Compare the two and say which tiers would pay for themselves on my numbers.
- If I wrote "skip" or gave no hours lost, do no money sums and say the verdict rests on the limits alone.
- Never put a number on credit prices, message counts or my time that I did not give you or that FACTS TO USE does not contain; write NOT STATED.

Job 4. Give one verdict.
- If my log has no plan-limit hits (Groups A to D), the verdict is STAY ON PLUS.
- When the employer-seat case under SOMETHING ELSE and TRY PRO both fit, lead with SOMETHING ELSE and give TRY PRO as the next step if the seat's limits would still block me.
- STAY ON PLUS: most hits are Group D or E, or the cheaper fixes in Job 2 other than credits would remove most of them, or the break-even hours are well above what I lose. List the fixes in the order to try them.
- TRY PRO 100 FOR ONE MONTH: Group A hits on Extra High or Pro thinking that my work needs, Group A limits on Medium or High that still block paid work most days after trying a lower level, or Group B hits most working days that the Job 2 fixes would not remove. Pro 200 only if my log shows I would run through Pro 100's weekly allowance, which you cannot know from a Plus log, so say that and default to Pro 100. Pro 500 only if I need Ultrafast, and warn me it uses the allowance at 8 times the rate.
- SOMETHING ELSE: if the work is for my employer and it offers Business or Enterprise, tell me to ask for a seat, and say standard Business seats get the same Work and Codex estimates as Plus, so it moves the bill more than the limit, while a Business Premium seat has no five-hour limit. If I already pay for another AI tool, suggest moving the kind of task that hits limits most to it for a week as a test, without claiming anything about its limits. If credits fit Group B, say so.
- Never recommend a tier above the most I said I will spend. If my ceiling is below the Pro 100 price my checkout shows (US$100 if I did not say), the answer is STAY ON PLUS or SOMETHING ELSE.

Job 5. If the verdict is TRY PRO, give me the month as a short plan.
1. Before upgrading, note my usage page readings as a baseline.
2. Upgrade, then schedule the cancellation the same day, so it ends unless I decide to keep it. Tell me to write a reminder for two days before the renewal date shown under Settings, then My Plan.
3. Keep the same log for the month, on the same kind of work.
4. On the calendar date, compare: did my top-ranked hits stop, and was the work they unblocked worth the extra cost from Job 3? If yes, undo the cancellation. If not, let it end.

MY RESULT
End with: the group counts, the top cost, the arithmetic, the verdict in one line, the cheaper fixes to try, and the cancel date if any. Tell me to save this answer somewhere, since you may not have this chat later.

SELF-CHECK BEFORE YOU ANSWER
- Every hit is sorted with its message quoted, or marked NOT STATED.
- Every fact you used is in FACTS TO USE; anything else is marked NOT STATED.
- Redo each sum once and confirm it matches.
- The verdict is within my spending ceiling and follows the rules in Job 4.
- You gave no advice beyond whether this subscription is worth it for my use.

What a log and the sum look like

One log line: Tue 2:10pm | Work | Astra in Work, Fast on | [the exact limit message, with its reset time] | waited 3 hours, report went out late

What the prompt does with it: with Fast on, its first fix is turning Fast off. This is one sample line; the prompt wants at least 3 days of them.

The sum: if the hits keep coming after Fast is off, and an hour of your time is worth US$40, Pro 100’s extra US$80 a month (US$100 minus US$20) pays for itself once it saves you 2 hours a month.

What Pro changes, limit by limit

Thinking in Chat

On Plus: Medium and High thinking only.

On Pro: adds Extra High and Pro thinking, run by GPT-6 Pro. It still has weekly limits, and on Pro 100 the two Pro models share one weekly allowance.

Verdict: worth it if your work needs the top thinking levels; for routine questions, a lower level is the cheaper fix.

Work and Codex

On Plus: OpenAI estimates 5 to 45 local messages per five hours on GPT-6 Astra, and 15 to 160 on GPT-6.1 Sol, with weekly limits on top, and Plus gets only limited Astra.

On Pro: no five-hour limit for now, though weekly limits may apply, and OpenAI warns a larger allowance does not guarantee twice as many finished tasks.

Verdict: the strongest reason to upgrade. Try the cheaper fixes first: turn off Fast mode, which burns the allowance at 2.5 times the rate, or switch routine tasks to GPT-6.1 Sol. If you only hit the wall a day or two a week, buy credits under Settings > Usage, if your account offers them.

Long files and long chats

On Plus: about 320 pages of text with a reasoning model, about 40 with Instant.

On Pro: about 680 pages with reasoning, about 250 with Instant.

Verdict: only matters if your files are longer than that and cannot be split.

Pro-only extras

What you get: Dots, ChatGPT Space, and monthly API credits of US$100, US$200 or US$500 by tier, which expire at the end of each billing cycle. Ultrafast is on Pro 500 only and uses the allowance at 8 times the rate.

Verdict: a bonus if you would use them, and a weak reason on its own.

Try Pro for one month without getting stuck

Pro is billed monthly only, so a one-month trial costs one month’s fee.

  1. Note your usage first. Open Settings > Usage or the Codex usage dashboard and save the readings, so the month has a baseline.
  2. Pick Pro 100. A Plus log cannot show whether you would outgrow it, so start at the lowest tier and move up only if its weekly limits block you.
  3. Schedule the cancellation the same day. OpenAI says you keep access to the end of the billing period and can undo it before then. Cancel where you paid: ChatGPT, Apple or Google Play.
  4. Set a reminder two days before renewal. Run the prompt again with the month’s log, then undo the cancellation only if the work Pro unblocked was worth the extra cost.

The honest bit

  • Changing in December: OpenAI’s business pricing page says that starting in December, plan usage will share one limit across all of ChatGPT and Codex (read 13 November 2026). Run the prompt again on a December week before you renew.
  • No fixed numbers: OpenAI publishes estimates for Work and Codex on Plus, and none for Pro’s weekly limits or for most Chat limits, so nobody can tell you in advance how far Pro stretches for your work.
  • Your check: after a full week on Pro, compare Settings > Usage with your Plus baseline on the same work. If you hit a Pro weekly limit and wait for the reset most weeks, Pro 100 is too small; test Pro 200 next if it fits your budget, once Fast mode is off.

The facts here come from OpenAI’s pricing page, its help pages on Pro tiers, models, Work and Codex, and credits, and its Work and Codex pricing guide, read 12 November 2026.

Do it now

Start a note today and add every limit message you hit this week, then paste it into the prompt on Friday.

A few quick questions

Can I move between Pro 100, 200 and 500 later?

OpenAI says Settings > My Plan shows your subscription and the plan changes open to you. If your tier changes sooner than you expected, compare it there with the change you confirmed, and if you paid through Apple or Google Play, check that store account too.

Can OpenAI support reset my limit if I run out early?

OpenAI says Support cannot reset ChatGPT or Codex limits, though it can look into usage that was counted wrongly. Hitting a limit does not restrict your account either; wait for the reset time ChatGPT shows, or switch to another model.