ChatGPT "reached concurrency limit": what it means and 7 fixes in order

In brief

ChatGPT's "Reached concurrency limit" error means too many requests were running at the same moment, not that you spent your plan's allowance. As of November 2026 OpenAI publishes no concurrency number for any plan. The same wording also appears during OpenAI outages. Close extra tabs, press Stop, wait a minute, then send one message.

What does "ChatGPT reached concurrency limit" mean?

It means ChatGPT refused a new request because too many were already in progress on your account at that moment. The limit counts what is running now, not what you used today, which is why it can show up on your first message of the morning. A reply that is still streaming, a second tab, a phone you forgot about and a browser extension all count.

A brushed silver OpenAI knot on a black plinth in a dark studio, with four glowing amber streams running from small chrome browser windows into one narrow chrome funnel, and one stream held back by a gate.

That gap is the real finding. OpenAI's rate limit guide measures requests per minute (RPM), requests per day (RPD), tokens per minute (TPM), tokens per day (TPD) and images per minute (IPM), and says a limit can trip on whichever is reached first. It never mentions concurrency. So the message in the ChatGPT app comes from a rule OpenAI hasn't written down, and nobody outside OpenAI can tell you the threshold. A page that gives you one is guessing.

OpenAI documents five rate meters. None of them counts simultaneous requests
Meters OpenAI documentsrequests and tokens,per minute and per day,plus images per minute5 metersConcurrencynumbers publishedfor any plan, in the guideor the error codes page0 numberseach square is one meter in OpenAI'srate limit guide, read 14 November 2026
What you seeWhere it showsWho reports itFirst move
Reached concurrency limitChatGPT web and appThird-party guides only (aifreeapi, IPWeb)Stop the running reply, close extra tabs, wait a minute
Too many concurrent requestsChatGPT web and appNewsweek (10 June 2025 outage report) and the OpenAI community forumSame as above, then check status.openai.com
Failed to load subscription: Too many concurrent requestsSettings and account screensA community.openai.com post of 31 January 2025. The longer ending "Contact support@openai.com if error persists." is from aifreeapiReload once with only one tab open
Unable to load GPT: too many concurrent requestsCustom GPT pagesaifreeapi onlyClose other chats, then reopen the GPT
Your current concurrent conversation limit has been reached. Please try again later.Older wordingaifreeapi onlySame as above

OpenAI doesn't document any of those five strings, as far as we could find. Only "Too many concurrent requests" has a source outside the troubleshooting blogs. Newsweek quoted it during the 10 June 2025 outage, and a forum user posted the subscription version on 31 January 2025. The one reply in that thread guessed at an unreported outage and no cause was given. Treat the other wordings as reported by one third party. Match the sense, not the letters.

That outage coverage matters for what you do next. The same message was on screen while OpenAI's status page said some users were seeing "elevated error rates and latency" and that it was investigating. So the message can come from OpenAI's side, with nothing wrong in your tabs.

How long do you have to wait?

Wait 30 to 60 seconds after you stop everything that was running, then send one message. No OpenAI page gives a duration for this error. aifreeapi and izzedo both suggest 30 to 60 seconds, while StatsTable says 1 to 5 minutes, and none of them cites OpenAI. Treat the range as a working rule, not a promise. aifreeapi's guide, a third party, says the help center had no article on these messages as of 1 October 2026.

A different message does name a wait. "You are making requests too fast! Please wait ~10 seconds" is reported with a ten second delay, and a "temporarily limited access" variant is reported with a few minutes. Those are speed limits on how fast you send, which is a separate thing from how many replies run at once. If a message names a time, believe it. If it names none, don't sit out an hour: the overlap is usually gone in seconds.

Two things stretch the wait. The first is an OpenAI incident, because then the limit has nothing to do with you and clears when they fix it. The second is a loop of resending. OpenAI's rate limit guide says failed attempts still count toward your per minute limit, so hammering Send keeps the counter up. That guide is about the API, but the habit hurts in the app too.

What fixes it, in order?

Work down this list and stop at the first fix that works. The order runs from the most common cause, overlapping requests, to the least common, a broken session or an OpenAI fault. Neither OpenAI nor the guides we read publish how often each cause happens, so the order is a judgment, not a measured ranking.

  1. Stop the running reply and wait 30 to 60 seconds. Press Stop generating on anything still streaming. Don't regenerate more than once.
  2. Close every other ChatGPT tab, window and device. Two phones and a laptop generating at once is the textbook case. If someone else could be using your login, sign out of all devices and sign back in on one.
  3. Switch off whatever sends prompts for you. Extensions, scripts and automations count as extra requests. Let uploads and long tasks finish, or cancel them.
  4. Send one short plain text message. "Reply with OK" is enough. If it works, bring your real work back one request at a time.
  5. Check status.openai.com. It lists ChatGPT components such as Conversations, File uploads and Login. An open incident means you wait it out.
  6. Try a private window, then clear site data. If private works and normal doesn't, sign out, clear site data for chatgpt.com, restart the browser and sign in again. Change one variable at a time: extensions, VPN, network, device.
  7. Contact OpenAI. If it persists across browsers, devices and networks with no incident, report it at help.openai.com. Include the exact message, the model, the conversation URL and the time.

If the error comes on your very first message with nothing else open, the cause is more likely an incident or a stale session than anything you did. That's the case for steps 5 and 6. If a different error is on your screen, ChatGPT too many requests and ChatGPT network error cover the two lookalikes, and what to use when ChatGPT is down covers a real outage.

Does paying for Plus or Pro raise the concurrency limit?

No source we found says it does. OpenAI publishes no concurrency number for Free, Go, Plus, Pro or Business, so there is nothing to compare, and a paid plan isn't protected from incidents. Upgrading to clear this one message is a gamble. If you pay already and still see it, the fixes above are the same.

The cap that paying does move is on features and models, which is a separate problem and the point where a second place to work starts to help. ChatGPT limit reached sorts the three kinds of limit message apart, and is ChatGPT Plus worth it does the math on upgrading.

If you came here because ChatGPT keeps stalling your work and a second tool would help, Whizi is one. Its free plan gives 10 messages a day on Whizi AI, Auto and Claude Haiku 5.5, 2 pictures a month and 5 photo uploads a day, with no card needed. Pro carries Claude, GPT and Gemini, and its first week is $1.99. This error belongs to ChatGPT, so Whizi can't clear it. It gives you somewhere to keep working while it clears.

Is it the API's 429 error instead?

If you see it in code or a tool that calls OpenAI, it is probably an HTTP 429, and that is a different system with documented rules. A ChatGPT subscription doesn't change API limits, and the API bills separately. OpenAI's error codes page is the one place that uses the word. Under "Rate limit reached for requests" it lists "a loop or a script that makes frequent or concurrent requests" as a cause, with no number attached. It lists the 429 causes this way:

429 heading on OpenAI's pageWhat OpenAI saysThe fixDoes retrying help?
Rate limit reached for requestsYou "have submitted too many tokens or requests in a short period of time"Pace requests, follow Retry-After, back off with jitterYes, after a pause
Slow downTraffic rose faster than the service can handle, even inside your limitsRamp traffic up graduallyYes, after a pause
Credit balance exhaustedYour "prepaid credit balance is depleted"Add credits in billing settingsNo
Organization or project spend limit reachedA spend or usage limit was hitRaise or fix the limitNo

OpenAI's rate limit guide says to treat the Retry-After header as a minimum wait: "Treat this value as a minimum." It adds a random delay, caps the number of attempts, and warns against nested retry loops with SDK retries. For fast growth it says that after you reach 1M input tokens per minute you should not raise traffic by more than 50% every 15 minutes. It also names rate_limit_error with the code slow_down for a sudden jump in traffic, and a 503 with server_is_overloaded when the model itself is overloaded. None of these is the app's concurrency message, which is why a guide that mixes them up sends you after the wrong fix.

One practical test: the ChatGPT message alone doesn't prove an HTTP status. Treat it as an API 429 only when you have the actual response and its error details.

Workflow checklist
  • Press Stop generating and wait 30 to 60 seconds before sending again.
  • Close every other ChatGPT tab, window and phone, and keep one open.
  • Turn off extensions, scripts and automations that call ChatGPT.
  • Send one plain text line before you resume real work.
  • Check status.openai.com for an open ChatGPT incident.
  • If you are in code, read the full 429 response before changing anything.
Common questions

Frequently asked questions

What does "ChatGPT reached concurrency limit" mean?

It means too many requests were running at once on your account, not that you used up a daily or monthly allowance. A reply still streaming, a second tab, another device or an extension can each add one. OpenAI documents no concurrency number and, per aifreeapi, had no help article for it as of 1 October 2026.

How long does the ChatGPT concurrency limit last?

No OpenAI page gives a duration. Troubleshooting guides suggest waiting 30 to 60 seconds after you stop everything that was running. If an OpenAI incident is open, it lasts as long as the incident does. A message that names its own wait, such as about 10 seconds, should be believed.

Does ChatGPT Plus fix the concurrency limit?

No source we found says it does. OpenAI publishes no concurrency threshold for any plan, and paid accounts are hit by incidents too. Try the fixes in order before paying, because overlapping tabs and extensions cause the error on every plan.

Is "reached concurrency limit" the same as the API's 429 error?

No. The app message counts requests running at once and is undocumented. The API's 429 is documented: the rate limit variant says you submitted too many tokens or requests in a short period, and separate credit and spend variants cover billing. Retrying with backoff helps the first and never helps the second.

Can ChatGPT show "too many concurrent requests" when nothing is wrong on my side?

Yes. Newsweek quoted the message during the 10 June 2025 outage, while OpenAI's status page reported elevated error rates. If it appears on your first message with one tab open, check status.openai.com before changing anything.