If ChatGPT tells you “Context window full” or “Chat memory full”, the fastest fix is to click New chat in the sidebar and paste in only the part you are still working on. Deleting saved memories will not help. OpenAI says both notices mean the current conversation has used up the space ChatGPT can work with, and that the limit applies to that one chat, not to your account. The harder case is when a brand-new chat shows the same notice after a handful of short messages. Many people have reported exactly that in early November 2026, and OpenAI’s Help Center now has a specific checklist for it. This guide walks through both cases in plain steps, explains how big a chat is on each plan, and lists what to tell OpenAI Support if nothing works.

Everything below was checked on November 8, 2026 (Beijing time) against three official OpenAI pages: the Help Center article Why does ChatGPT say “Context window full” or “Chat memory full”?, the ChatGPT Free Tier FAQ, and the plan comparison on ChatGPT Plans and Pricing. The Help Center article on Temporary chat in ChatGPT is used for the Temporary chat steps. User reports come from the OpenAI Developer Community forum and are labeled as reports, not confirmed facts.

Key facts:

• What the notice means: “Context window full” and “Chat memory full” both refer to how much information ChatGPT can use in the current conversation. Your messages and ChatGPT’s replies both take up space, so “there’s no fixed number of messages per chat” (Free Tier FAQ).
• What does not help: “You don’t need to turn off Memory, delete saved memories, or clear other chats” (Free Tier FAQ). “Deleting saved memories won’t make more room in the chat” (Help Center).
• You do not have to wait or pay: you can start a new chat right away on the Free plan. The notice gives no reset time for the original chat; an upgrade option, if shown, lets you continue in that same chat (Free Tier FAQ).
• Chat size by plan (GPT Instant): 27K on Free, 54K on Go, 54K on Plus, 128K on Pro. Thinking (GPT Reasoning) chats: “Varies” on Free, 256K on Go and Plus, 400K on Pro (pricing page).
• New chat already full? Send one short test message in a separate new chat, check OpenAI Status, then try a private browser window. If it keeps happening, contact Support with specific details (Help Center).

What “context window” actually means

Think of every ChatGPT conversation as a desk with limited space. Everything ChatGPT needs to answer you has to fit on that desk at once: every message you sent in this chat, every reply it wrote back, any text from files you pasted or uploaded, plus some things you never see.

OpenAI’s pricing page spells out the hidden part in a footnote. ChatGPT “manages a shared context window to understand your request, track the conversation, retrieve relevant information, and generate responses.” The part available for your input “is smaller than the total window, as space is also used for system instructions (including tools and personality), memories (if enabled), and internal processing (reviewing information, reasoning, and response generation).” It adds that the space for user input “is an approximation and may change dynamically based on features in use and any memory content” (ChatGPT pricing).

So two chats with the same number of messages can fill up at very different speeds. The Free Tier FAQ puts it simply: “A few long exchanges can take up more room than many short ones, so there’s no fixed number of messages per chat” (Free Tier FAQ). Pasting a long report, asking for a 3,000-word draft three times, or uploading a big file each eats a lot of space in one go.

Two things the notice is not:

• It is not your saved Memory being full. Saved memories (the facts ChatGPT remembers about you across chats, managed under Settings) are a separate feature. OpenAI says deleting them “won’t make more room in the chat” (Help Center).
• It is not a daily message cap. Free users have “unlimited everyday text chats, subject to abuse-prevention safeguards”, and tools such as file uploads, image creation and data analysis have separate usage limits that show their own notices (Free Tier FAQ). If you see a message about waiting until a certain time, that is a usage limit, not the context window. (Paid-plan usage limits are changing in December; see our guide to the shared ChatGPT and Codex usage limit.)

How big is a chat on each plan?

The ChatGPT pricing page lists two sizes per plan: one for the everyday fast model (GPT Instant) and one for the thinking models (GPT Reasoning). “K” means thousands of tokens; a token is a small chunk of text, roughly a short word or part of a word. OpenAI also translates the input maximum into approximate pages of text.

Plan	GPT Instant total context	GPT Instant input maximum	GPT Reasoning total context	GPT Reasoning input maximum
Free	27K	~12 pages of text	Varies	Varies
Go	54K	~40 pages of text	256K	~320 pages of text
Plus	54K	~40 pages of text	256K	~320 pages of text
Pro	128K	~250 pages of text	400K	~680 pages of text

Source: ChatGPT Plans and Pricing, checked November 8, 2026. OpenAI marks the input maximums as approximations that can change.

What this means in practice:

• On Free, an Instant chat holds about 12 pages of your text, total. That includes ChatGPT’s replies to you. A long editing session on one document can fill it surprisingly fast.
• Go and Plus have the same Instant size (54K). Upgrading from Go to Plus does not make an Instant chat longer. It is the thinking models (256K) where both plans get much more room.
• Switching to a thinking model can give you more room on paid plans, because the Reasoning window is larger. The trade-off is that thinking models count more heavily against paid usage limits, especially after December’s shared allowance change.
• Pro is the only plan with a much larger Instant window (128K), so heavy single-chat users on Plus will not escape the notice just by staying on Instant.

Fix 1: The chat really is full (the normal case)

This is the common situation: you have been in one conversation for a long time, or pasted large chunks of text, and ChatGPT now refuses to continue. Here is the order OpenAI recommends, with what each option costs you.

Step 1. Decide whether you need the old context at all

Ask yourself what you still need from the old chat. Usually it is one thing: the latest version of a draft, a list of decisions, or a few key instructions (“write in British English”, “keep it under 200 words”). Everything else can stay behind.

Step 2. Use the summary option if the notice offers one

OpenAI says that if the notice offers to start a new chat with a summary, “you can use it to pick up where you left off.” It warns that “a summary may leave out details” (Help Center). The Free Tier FAQ is more specific: the summary “may leave out an earlier instruction or the exact wording of something you wrote” (Free Tier FAQ).

So after you continue with a summary, do a quick check:

1. Read the summary ChatGPT starts with.
2. Re-paste any rule that matters and is missing, in one short message: “Reminder: keep the tone friendly, use bullet points, under 300 words.”
3. If you are editing text, paste the exact current version rather than trusting the summary’s paraphrase.

Step 3. Or start a clean chat and bring only what you need

You can “start a new chat right away for free” (Help Center). The old chat does not disappear. OpenAI says “you can still read earlier messages even when you can’t send another one in that chat” (Free Tier FAQ).

1. Open the old chat from your history and copy what you need, for example the paragraph you are editing.
2. Click New chat in the sidebar.
3. Paste only that material, plus one line saying where to pick up: “Here is the latest draft of my cover letter. Please tighten the second paragraph.”
4. Add older instructions only when they become relevant again.

OpenAI’s own advice for drafts is exactly this: “paste the passage you want to revise and tell ChatGPT where you’d like to pick up” (Free Tier FAQ).

Step 4. Upgrade only if you need that exact chat

If the notice offers an upgrade, choosing it lets you continue in the original chat (Help Center). Look at the table above before paying, though. If you mostly use Instant, Go and Plus both give a 54K window, double Free’s 27K, but not unlimited. A chat that filled up on Free in an hour may still fill up on Plus in an afternoon.

Can’t find or open the old chat?

• Not in the sidebar: search your chat history for a phrase you remember (Free Tier FAQ).
• Opens blank or will not load: refresh the page or restart the app and reopen it from history. If it still will not load, open a private browser window, sign in to the same account, and try there (Free Tier FAQ).

Fix 2: A brand-new chat says it is full after a few messages

This is the frustrating case, and it is what many people are hitting right now. On the OpenAI Developer Community forum, users reported in early November 2026 that new chats were showing “Context window is full” after only 10 to 30 messages, on the web and in the Android and iPhone apps. One user reported the problem first appeared for them on October 20, went away the next day, and returned on November 5. Others reported that both Free and paid (Go and Plus) accounts were affected (forum thread). These are individual user reports. OpenAI has not published a root cause, and replies from OpenAI Support that users posted on the forum say Support could not confirm a technical defect, a change to the context limit, or a resolution date (forum thread).

What OpenAI has published is a short troubleshooting sequence in the Help Center. Do these in order:

Step 1. Make sure you are really in a separate new chat

If you used the “continue with summary” option and hit the limit again, that new chat starts with carried-over context. OpenAI says to “choose New chat in the sidebar” instead (Help Center). A support reply posted on the forum explains why: “Continue in new chat” can carry information over, which reduces the space available (forum thread).

Step 2. Send one short test message

In that separate new chat, send something tiny. OpenAI’s own example is “What’s the capital of France?” (Help Center). This tells you whether the problem is your content or something on OpenAI’s side:

• The short message works, but your real task fills up quickly: your task is large. Cut what you paste in, split the work across several chats, or use a thinking model on a paid plan (larger window).
• Even the short message triggers the notice: this is not your content. Go to Step 3.

Step 3. Check OpenAI Status

Open status.openai.com. OpenAI says that if there is an incident affecting ChatGPT conversations, try again after it is resolved (Help Center). Nothing you change in your settings will beat a server-side incident, so this check can save you an hour of fiddling.

Step 4. Try a private browser window

If there is no related incident, OpenAI’s next step is to sign in to the same account in a private (incognito) browser window and send the same short message in a new chat (Help Center). This rules out a stuck browser session, extensions, or cached data. Do this even if you normally use the phone app, because it gives Support a useful comparison.

Step 5 (optional). Try an Unpersonalized Temporary chat

Support replies posted by forum users also suggested a Temporary chat with personalization turned off (forum thread). OpenAI lists this as a workaround to try, not a guaranteed fix. The reason it can help: an Unpersonalized temporary chat “does not use memory, custom instructions, or plugins” (Temporary chat help), so less hidden material competes for space. How to start one:

1. Open a new chat.
2. Select Temporary.
3. Choose Unpersonalized before you send your first message. You cannot change this after the conversation starts.
4. Type your message.

Two things to know: a temporary chat stays out of your history unless you save it, and OpenAI may keep a copy for up to 30 days for safety purposes (Temporary chat help). If you want to keep the work, save it as a regular chat before you leave. Saving converts it into a normal chat that follows your usual personalization settings. If privacy is the reason you are reading this, our guide on stopping ChatGPT and Codex from training on your data covers the related settings.

Step 6. Contact Support with the right details

If none of the above works, OpenAI asks you to contact Support, “even if it only happens in the app.” Include (Help Center):

• your app or browser version,
• the time and your time zone,
• about how many messages you sent,
• what happened when you tried the private browser window,
• optionally, a link to a test chat you are comfortable sharing (you can keep your original conversation private).

Write this down as you go through Steps 2 to 4 so you only have to describe it once.

Things that waste time (and what to do instead)

Common attempt	Does it help?	Why
Deleting saved memories	No	OpenAI: deleting saved memories “won’t make more room in the chat” (Help Center)
Turning Memory off or clearing other chats	No	The limit applies to the current conversation only (Free Tier FAQ)
Waiting for a reset	No	The notice “doesn’t give a reset time for the original conversation” (Free Tier FAQ)
Clicking “continue with summary” again and again	Partly	Each summary carries context over and may drop details; start a separate New chat instead
Creating a new account	Not recommended	Forum users report it works only temporarily, if at all (forum thread), and you lose your history
Upgrading from Go to Plus for longer Instant chats	No	Both have a 54K Instant window (pricing)
Starting a New chat with only the text you need	Yes	This is OpenAI’s main recommendation

Habits that keep chats from filling up

These are practical habits that follow from how OpenAI describes the context window. None of them are special settings, just ways of working:

1. One task per chat. Start a fresh chat when you switch from “plan my trip” to “fix this email”. Long mixed-topic chats fill up fastest and also confuse the model.
2. Paste the part, not the whole. If you only need help with section 3 of a report, paste section 3. Every pasted page counts against the window.
3. Ask for shorter replies when you are iterating. “Only show the changed sentences” uses far less space than reprinting a full draft each round. Replies count too.
4. Keep a running “brief” you can paste. Maintain a short note with your key instructions (tone, length, audience) in a notes app. When you start a new chat, paste it first.
5. Use the right model for long documents on paid plans. On Go, Plus and Pro, the thinking models have a much larger window (256K to 400K) than Instant (54K to 128K) (pricing). Keep in mind they draw more from paid usage allowances.
6. If you hit this a lot, compare other assistants for long single-document work. Different apps have different chat sizes and limits. If you want to compare, see our guide to using Gemini 3.6 Flash.

FAQ

Is “Chat memory full” the same as “Memory is full” in settings? No. “Context window full” and “Chat memory full” both describe the current conversation (Free Tier FAQ). Saved memories are a separate feature, and OpenAI says you do not need to delete them to fix this notice.

How many messages can I send before a chat fills up? There is no fixed number. It depends on how long the messages and replies are (Free Tier FAQ). As a rough guide, a Free Instant chat’s input maximum is about 12 pages of text (pricing).

Do I lose my old chat? No. You can still read it, copy from it, and find it in history. You just cannot send new messages in it unless you upgrade (if offered).

Does this happen on paid plans? OpenAI’s Help Center article is written for the Free plan, but paid plans also have finite windows (see the table). Forum users have also reported the early “new chat is full” behavior on Go and Plus (forum thread). A support reply posted on the forum said upgrading “is also not guaranteed to resolve the specific premature context-window behavior” (forum thread).

Is OpenAI fixing the “new chat already full” problem? As of November 8, 2026, OpenAI has not published a root cause or a fix date. Its Help Center article was updated in the past day with the troubleshooting steps above. Check OpenAI Status for any incident notice.

Does a Temporary chat fix it for sure? No. It is a reasonable thing to try because an Unpersonalized temporary chat does not load memories, custom instructions or plugins, but OpenAI presents it as a workaround, not a confirmed fix.

Bottom line

For the everyday case, treat the notice as “this conversation is done”: copy what you need, click New chat, paste only that, and keep going. Do not delete memories and do not wait for a reset, because neither helps. If a brand-new chat fills up after one short message, follow OpenAI’s short checklist (separate new chat, test message, OpenAI Status, private window), then send Support the details it asks for. And before you pay to upgrade just to get longer chats, check the table: on Instant, Go and Plus give you the same 54K window.