ChatGPT Usage Limits (Oct 2026): What's Unlimited, 5-Hour Reset
Jonhisking · Updated October 19, 2026 · Updated October 19, 2026

"You've reached your limit" used to be the most common complaint about ChatGPT, and it still shows up in 2026, just in different places. OpenAI removed the cap on everyday text chat, then added GPT-6, GPT-6 Astra, a shared allowance for ChatGPT Work and Codex, three Pro tiers and a five-hour window that resets on its own schedule. The result is that "how much can I use ChatGPT?" now has several answers depending on what you are doing. This guide to ChatGPT usage limits in 2026 goes through every plan and every meter, how each one resets, what makes the allowance run out faster, and what you can do when you hit a wall. All numbers come from OpenAI's help center, pricing page and Codex documentation, checked on October 19, 2026.
🧮 Open the Subscription vs API calculator →
Is there a limit to how much you can use ChatGPT?
Yes, but not where most people expect. There are now three separate kinds of limits, and they do not draw from each other:
- Chat limits. Everyday text chat is unlimited on every consumer plan, subject to abuse guardrails. What is limited inside Chat is the higher thinking levels, the Pro models (GPT-6 Pro and GPT-5.6 Sol Pro) and tools such as file uploads, image generation, data analysis and deep research.
- The Work and Codex allowance. ChatGPT Work and Codex share one included allowance per plan. On Plus and Business Standard it runs on a five-hour window plus a weekly window. This is where GPT-6 Astra and GPT-6.1 Sol live, and it is the meter most heavy users hit.
- Voice and other features. ChatGPT Voice (GPT-Live) has daily hours per plan, and some features have their own allowances.
Hitting one limit does not lock you out of the others. If your Work allowance is used up, you can usually still chat normally; if GPT-6 Pro hits its weekly limit, ChatGPT can fall back to another model.
What changed in October 2026
OpenAI's release notes for October 1–16 add no new caps, but four changes affect what you see:
- GPT-6 in Chat on every plan (October 7–8). Plus, Pro, Business and Enterprise get GPT-6 Sol; Free and Go get GPT-6 Luna from October 8. The new Intelligent UI, which can put charts and small tools inside an answer, has no separate quota: your plan's usual limits apply. OpenAI's models help page still lists GPT-5.6 Luna for Free and Go, including Think, so an account that has not been updated yet may still show it.
- Codex message predictions (October 9, beta, personal Pro). Generating suggested next messages does not count toward Codex limits or credits. The messages you send still do.
- Sites (October 15, Plus and Pro). There is no fixed limit on the number of Sites, but usage and storage limits still apply.
- GPT-6 Pro update (October 16, Pro). Faster answers and parallel agents in the Chat tab. The release note does not mention any change to the weekly GPT-6 Pro allowance.
ChatGPT usage limits by plan

| Plan | Price per month | Everyday chat | Thinking levels in Chat | GPT-6 Pro in Chat | Work and Codex | Five-hour window |
|---|---|---|---|---|---|---|
| Free | $0 | Unlimited* | Think (GPT-6 Luna) | – | Limited, desktop app | – |
| Go | $8 | Unlimited* | Think (GPT-6 Luna) | – | Limited, desktop app | – |
| Plus | $20 | Unlimited* | Medium, High | – | Included | Yes |
| Pro 100 | $100 | Unlimited* | Medium to Extra High, Pro | Yes, weekly allowance | Larger allowance | No |
| Pro 200 | $200 | Unlimited* | Medium to Extra High, Pro | Yes, weekly allowance | Larger allowance | No |
| Pro 500 | $500 | Unlimited* | Medium to Extra High, Pro | See note below | Largest allowance, Ultrafast | No |
| Business Standard | per seat | Unlimited* | Medium to Extra High, Pro | 15 Pro messages a month | Included | Yes |
| Business Premium | per seat | Unlimited* | Medium to Extra High, Pro | 50 Pro messages a week | 500% of Standard | No |
*Subject to abuse-prevention safeguards. Prices are US list prices before tax.
- "Unlimited" means everyday text chat only. OpenAI's pricing page lists everyday text chat as unlimited on Free, Go, Plus and Pro, and its Business help page says unlimited access applies to Instant chat and "does not make reasoning, Pro, ChatGPT Work, Codex, or other tools unlimited."
- GPT-6 Pro on Pro 500. OpenAI's help center lists GPT-6 Pro for Pro $100, Pro $200, Business and Enterprise; its Pro tiers page says Pro 500 "offers the highest included usage of the three plans." Check the model picker on your account.
- Pro usage tiers. OpenAI's help page only says Pro 200 includes more than Pro 100 and Pro 500 the most. The figures usually quoted (Pro 100 at 500% of Plus, new Pro 200 at 1,000%, Pro 500 at 2,500%) come from an OpenAI employee's post on X and press reports, not from OpenAI's pricing page. See ChatGPT Pro 100 vs 200 vs 500.
Everyday chat: unlimited, with guardrails
On August 6, 2026, OpenAI announced in its release notes that Free and Go users would get "unlimited text chats" from the following week, along with a Think button for harder questions. The pricing page now marks everyday text chat as unlimited on every consumer plan.
Three things are worth knowing about "unlimited":
- It covers ordinary text answers. On Free and Go, those come from GPT-6 Luna since the October 8 rollout (GPT-5.6 Luna on accounts that have not received it yet), and Think also uses Luna. On Plus and above, Instant uses GPT-6 Sol.
- Tools are still limited. OpenAI says separate limits "continue to apply to file uploads, image generation, voice, data analysis, and other tools." On Free and Go, the pricing page lists messages with uploads, image generation, deep research, memory and Codex as limited.
- Guardrails apply. Automated use, sharing an account or reselling access can lead to a temporary restriction. OpenAI says it will tell you when that happens.
So if you send normal text messages, you should not see a message cap on any plan. If you do, you are most likely using a thinking level, a Pro model or a tool.
Thinking levels and Pro models
The Thinking slider decides which model answers you, and that is where Chat limits actually bite.
| Thinking level | Model (GPT-6 selected) | Plus | Pro | Business | Free and Go |
|---|---|---|---|---|---|
| Instant | GPT-6 Sol | ✓ | ✓ | ✓ | GPT-6 Luna |
| Medium, High | GPT-6 Sol | ✓ | ✓ | ✓ | – |
| Extra High | GPT-6 Sol | – | ✓ | ✓ | – |
| Pro | GPT-6 Pro (powered by GPT-6 Astra) | – | ✓ | ✓ | – |
- Thinking limits depend on your plan. If you reach one, "ChatGPT may continue with another available model," and it shows when the allowance resets when that information is available.
- GPT-6 Pro is never unlimited. OpenAI states plainly that "a Pro subscription does not include unlimited use of the GPT-6 Pro model."
- Pro $100: GPT-6 Pro and GPT-5.6 Sol Pro draw from one weekly allowance, so switching between them does not give you more messages.
- Pro $200: when you reach the GPT-6 Pro weekly limit, ChatGPT switches automatically to GPT-5.6 Thinking at Medium. GPT-5.6 Sol Pro stays available while its daily allowance lasts.
- Business: Standard seats get 15 Pro messages a month and Premium seats 50 a week, shared across GPT-6 Pro and GPT-5.6 Sol Pro.
- Plus does not include GPT-6 Pro in Chat. Plus users get GPT-6 Astra in Work and Codex instead, under the shared allowance below.
Work and Codex: the shared allowance

This is the limit people mean when they talk about the "five-hour limit." ChatGPT Work and Codex share one allowance, and it does not correspond to a fixed number of messages. OpenAI gives ranges of local messages per five-hour window for Plus and Business Standard:
| Model | Plus | Business Standard |
|---|---|---|
| GPT-6 Astra | 5–45 | 5–45 |
| GPT-6.1 Sol | 15–160 | 15–160 |
| GPT-6 Sol | 15–150 | 15–150 |
| GPT-6 Luna | 350–3,000 | 350–3,000 |
| GPT-5.6 Sol | not listed | 10–100 |
| GPT-5.6 Terra | not listed | 25–200 |
| GPT-5.6 Luna | not listed | 250–2,000 |
The GPT-6 rows are from OpenAI's Codex pricing page and Astra help article. The GPT-5.6 rows are from the Business models and limits page, which also groups GPT-6.1 Sol and GPT-6 Sol together at 15–150 for Standard seats.
- The range is wide on purpose. A one-line fix and a long multi-step task are both "one message," but the second uses far more allowance. Cloud tasks can use more than local messages.
- Pro plans have no five-hour limit in Work and Codex for now, but they still have an included allowance that shows in Settings → Usage.
- Business Premium gets 500% of a Standard seat's included Work and Codex usage and no five-hour limit. Standard seats get "limited Astra usage" within their allowance; Premium seats can spend their full allowance on Astra.
- Free and Go get limited Codex and ChatGPT Work in the desktop app with GPT-6 Luna, according to the Codex pricing page.
- Switching models does not refill the pool. OpenAI says switching models "does not restore allowance in a shared usage pool." A smaller model only makes the remaining allowance go further.
The five-hour window, explained
Plus and Business Standard have two windows on Work and Codex at the same time:
- Five-hour window. It starts when you send your first Work or Codex message after the previous window ends. It is not a fixed clock time, so two people on the same plan can have windows that reset at different times.
- Weekly window. It caps how much included work you can do across the week.
You need allowance left in both to keep going. That is why you can hit the five-hour limit after an hour of heavy work even though your weekly meter still shows plenty left, and why a big weekend session can leave you blocked on Tuesday even with a fresh five-hour window.
Pro plans and Business Premium skip the five-hour window but still have an overall allowance with reset times. Free and Go do not show a five-hour window either: when we tested Codex on the free plan (see below), the usage menu showed a single monthly meter.
What makes the allowance run out faster
Prompt length alone is a poor guide. OpenAI lists model choice, context, reasoning, tool use, retrieval and caching as the main factors. The ones you control:
- Model. GPT-6 Astra uses the most per task; GPT-6 Luna the least. In the Plus ranges above, Luna gets 350–3,000 messages where Astra gets 5–45.
- Reasoning level. Higher effort uses more allowance and "does not always produce a better result." OpenAI even suggests trying Astra at Low or Medium before Sol at High.
- Speed modes. Fast mode uses your included allowance at 250% of the Standard rate. Ultrafast, available only on Pro 500 among Pro plans, uses it at 800% for GPT-6 Astra and GPT-6.1 Sol.
- Images. In Work and Codex, image generations use included limits about 300–500% as fast as similar turns without images, depending on quality and size.
- Context. Large instruction files, many connected MCP servers and long source material all add tokens to every message. OpenAI recommends trimming AGENTS.md files and turning off MCP servers you are not using.
- What one task actually uses. In our own measurements with Codex CLI 0.162 and GPT-6 Luna, Codex sent 14,678 input tokens before doing any work (a one-word reply). A small coding task, a brick-breaker game in one HTML file, used a median of 48,600 input tokens over 25 runs, most of it read from the cache. Adding 10 or 30 idle MCP tools did not change the 14,678: Codex loaded them only when a task needed them. The raw data is in Claude Code vs Codex tokens and prompt language and coding cost.
Voice, images and files
These features have their own allowances and are the most common reason Free and Go users still see limits.
| Feature | Free | Go | Plus | Pro |
|---|---|---|---|---|
| Voice (GPT-Live), per day | Limited | Up to 3 hours (GPT-Live-1 mini) | Up to 3 hours | Pro 100: up to 15 hours; Pro 200: unlimited |
| Image generation | Limited | Higher limit | Limits apply, higher quality | Unlimited* |
| File uploads | Limited | Higher limit | Expanded | At least Plus |
| Deep research | Limited | Limited | Expanded | Maximum |
Voice hours are from OpenAI's September 9, 2026 release notes; the rest is from the pricing page, which does not publish exact counts for most features. Plus and Pro no longer fall back to GPT-Live mini after reaching a voice limit. Voice in the desktop Codex app is different again: it uses your Codex usage budget at $0.05 per minute. If you generate a lot of images, the image cost calculator shows what the same images would cost through the API.
Context window: the "token limit" inside a chat

Separate from usage limits, every chat has a context window: the maximum amount of text the model can consider at once. When a conversation or document goes past it, older parts drop out of view, and quality often slips before that point.
| Free | Go | Plus | Pro | |
|---|---|---|---|---|
| Instant: total context window | 27K tokens | 54K tokens | 54K tokens | 128K tokens |
| Instant: approximate max input | ~12 pages | ~40 pages | ~40 pages | ~250 pages |
| Reasoning: total context window | Varies | 256K tokens | 256K tokens | 400K tokens |
The usable input is smaller than the total because the system instructions, memory and the model's own reasoning share the same space. Starting a new chat with a short summary is the simplest fix when a long conversation starts forgetting things. See Context windows explained and, for how many tokens your text uses, tokens per word.
How to check your usage
- ChatGPT: open Settings → Usage. It shows your remaining Work and Codex allowance, which window you are in and when each resets, plus any saved resets and your credit balance.
- Codex CLI: type
/statusduring a session to see your remaining limits. - In Chat: when a model allowance runs out, ChatGPT shows the reset time when that information is available.
OpenAI suggests checking the dashboard every week or two to learn your own pace.

This is the screen on a ChatGPT Free account in the Windows desktop app on October 13, 2026, where the menu is called Usage & billing. Three things stand out. It says Codex, Work, workspace agents and ChatGPT for Excel share one allowance, and that Chat conversations are not counted. Free shows a single monthly bar instead of five-hour and weekly meters; ours had 99% left with the reset 27 days away. And there is a switch that lets Codex spend a saved usage reset, but only when you ask for it in a conversation and less than 10% of your limit is left; resets are never used on their own. The Usage limit resets section below shows how many resets you hold (0 on this account) and what one restores: the five-hour limit, the weekly limit or both.
Which limit did I hit?
Because the meters are separate, the fastest way to fix a limit is to work out which one you reached. A quick guide:
| What you notice | The meter you probably hit | What to check |
|---|---|---|
| Work or Codex stops, but normal chat still works | Work and Codex five-hour window | Settings → Usage, reset time of the five-hour window |
| A fresh five-hour window, but Codex still says you are out | Work and Codex weekly window | Settings → Usage, weekly reset date |
| Pro thinking option greyed out or replaced by Medium | GPT-6 Pro allowance (weekly on Pro, monthly or weekly on Business) | Model picker and the reset time ChatGPT shows |
| High or Extra High answers switch to another model | Thinking limit for your plan | Reset time shown in Chat |
| Image, upload or deep research request refused | That tool's own allowance | Upgrade or wait; text chat still works |
| Voice ends after a few hours | Daily GPT-Live hours | Resets the next day |
| Everything stops and you see a notice about unusual activity | Abuse guardrails, not a usage limit | Contact Support if you think it is a mistake |
On a Business seat, the workspace's credit settings matter too: once the included allowance is used, extra Astra usage only continues if the workspace has credits and its spending controls allow it.
What happens when you hit a limit
You are not cut off mid-task. If you reach your limit during an active Work or Codex turn, the agent can finish that turn, subject to fair use. After that, your options depend on the plan:
- Wait for the reset. For the five-hour window that is at most a few hours; the reset time is in Settings → Usage.
- Use a saved (banked) reset. OpenAI sometimes grants resets, for example on September 3 and 4, 2026, during the Astra launch. A full banked reset refreshes both the five-hour and weekly limits and moves your weekly reset date.
- Buy an instant reset. Eligible personal Plus and Pro accounts can buy one in some countries. It refreshes both windows straight away but cannot be saved for later.
- Use credits. Plus and Pro users can buy credits to keep going in Codex, Work and the Office add-ins. Credits pay for usage; they do not unlock models such as Ultrafast on Pro 100 or 200.
- Use an API key for local work. Usage is then billed at API rates instead of your plan.
- Fall back in Chat. If a thinking or Pro limit is reached, choose another available model or let ChatGPT switch for you.
One thing that does not work: asking Support. OpenAI says Support "does not reset ChatGPT or Codex usage limits," though it will investigate if usage looks miscounted.
What we saw on the free plan
We ran Codex on a ChatGPT Free account for our Codex pricing test: the same brick-breaker game prompt, once on medium and once on extra high reasoning, with GPT-6 Luna. Both runs finished in 39 seconds. The usage menu had no five-hour or weekly meter, only one monthly bar, and after both games it showed 99% remaining with the reset a month away. Two small games used about 1% of a month's free allowance, which suggests that light Codex use fits comfortably on Free. A larger task on a real codebase reads more files and runs more steps, so it will use more.
Which plan avoids the limits you hit?
- You only chat and sometimes upload a file: Free or Go. Text chat is unlimited; Go raises the upload, image and voice limits.
- You hit limits on Thinking or image generation: Plus. It adds Medium and High thinking and expanded tool limits.
- You hit the five-hour Work or Codex window most days: Pro 100. It removes the five-hour window and enlarges the allowance.
- You run out of GPT-6 Pro or the Pro 100 allowance every week: Pro 200.
- You run agents all day or need Ultrafast: Pro 500, the only Pro tier with Ultrafast.
- You manage a team: Business Standard for normal use, Premium for people who live in Codex.
Plan or API?
If most of your use is chat, the unlimited text change makes the subscription hard to beat. If you use Codex or GPT-6 Astra heavily, compare the plan price with what the same work would cost on the API. GPT-6 Astra costs $10 per million input tokens and $50 per million output tokens on the API; see GPT-6 API pricing. Enter your own usage in the subscription vs API calculator, or use the coding agent calculator for Codex and Claude Code.
FAQ
What is the 5-hour usage limit on ChatGPT? It is the shorter of two windows on the Work and Codex allowance for Plus and Business Standard. It starts with your first Work or Codex message after the previous window ends and caps how much you can do in that window. A weekly window applies at the same time. Pro plans and Business Premium currently have no five-hour limit.
Does ChatGPT Free have a message limit? Not for everyday text chat. Since August 2026, Free and Go have unlimited text chats, subject to abuse guardrails. File uploads, image generation, voice, data analysis and other tools still have limits.
Do ChatGPT limits reset daily? It depends on the meter. Work and Codex use a rolling five-hour window and a weekly window; voice has daily hours; GPT-6 Pro has a weekly allowance on Pro plans. Settings → Usage shows the exact reset times for your account.
Does switching models get around the limit? In Work and Codex, no: all models draw from one shared pool, and switching does not restore it. A cheaper model only makes the rest last longer. In Chat, reaching a thinking or Pro limit can fall back to another model, which lets you keep working.
Is ChatGPT Plus worth $20 for the limits? If you only send text messages, Free or Go already gives you unlimited chat. Plus is worth it if you use the higher thinking levels, generate images, upload files often or want Work and Codex with GPT-6 Astra.
Which model does ChatGPT Free use now? GPT-6 Luna, rolling out to Free and Go from October 8, 2026. Plus and above get GPT-6 Sol in Chat. If your model menu does not show GPT-6 yet, your account is still on GPT-5.6 Luna.
Can OpenAI Support reset my limit? No. Support can investigate if usage was counted incorrectly, but it does not reset limits. Wait for the reset, use a saved reset, buy one if eligible, or use credits.
Limits change often. Check Settings → Usage for your own account, and OpenAI's help center for the current numbers.
Sources
- OpenAI Help Center: GPT-6 and other models in ChatGPT
- OpenAI Help Center: Managing usage with GPT-6 Astra in Work and Codex
- OpenAI Help Center: ChatGPT Business models and limits
- OpenAI Help Center: About ChatGPT Pro tiers
- OpenAI Help Center: ChatGPT release notes
- Notebookcheck: ChatGPT Free and Go users start getting GPT-6 Luna in Chat
- ChatGPT Learn: Codex and ChatGPT Work pricing and usage limits
- ChatGPT pricing
- OpenAI API: GPT-6 Astra model and pricing
- WinBuzzer: OpenAI Adds $500 ChatGPT Pro Plan, Cuts Allowance for New $200 Plan Subscribers
- Windows Report: OpenAI Launches $500 ChatGPT Pro 500 Plan
More articles
- Claude Haiku vs Opus for Coding: 330 Measured Runs
- Claude Code Cache Lasts 1 Hour: What a Break Costs (Measured)
- Indie Game Development Cost: Solo Dev Budget, Fees, Real $20 Case
- Grok API Pricing 2026: Grok 4.7, 4.3, Tools and Imagine Costs
- DeepSeek API Pricing 2026: Official Rates, Peak Hours, API Key
Enter how you use AI and see the monthly API cost next to every ChatGPT, Claude and Gemini plan.
Open the Subscription vs API calculator →