Don't stop here
Hand-picked guides our readers explore right after this one.
Master ChatGPT with advanced prompting techniques, mega-prompts, and proven frameworks
Read the guideUnlock Google's Gemini with multimodal prompting strategies
Read the guideCreate stunning AI images with Flux by Black Forest Labs using structured prompt techniques
Read the guide'You've hit your limit, please try again later' means your account has used up the messages allowed for that model inside a rolling time window. It is not an outage and not a bug on your device. The confusing part is the reset, because ChatGPT caps are rolling rather than calendar-based: the system counts how many messages you sent in the trailing window (commonly the last 3 or 5 hours depending on plan and model) and lets you send again as the oldest messages age out. That is why waiting sometimes frees up one message rather than the whole allowance. In 2026 ChatGPT runs on the GPT-5.5 family, with Free on GPT-5.5 Instant and a much smaller cap, Go and Plus on far higher caps plus a separate weekly Thinking allowance, and the Pro tiers ($100 and $200 per month) on effectively unlimited standard-model use. Reported numbers vary because OpenAI tightens caps during peak demand, so treat any published figure as approximate and check your own plan page for the current allowance.
'You've hit your limit, please try again later' appears when you send a message
The composer is greyed out or your message bounces back unsent
Answers silently switch to a smaller, faster model mid-session
A specific feature (images, file uploads, Thinking, voice) is blocked while normal chat still works
The limit clears one message at a time rather than all at once
You hit it much sooner than the number quoted in a blog post
The main cause. Each plan gets a set number of messages per trailing window, and heavy use in a short burst exhausts it. Free plans hit this fastest because the standard-model allowance is small before it downgrades you to a lighter model.
Image generation, file uploads, Thinking (deep reasoning), Deep Research, voice, and agent-style browsing all carry their own separate limits. You can be blocked on one while ordinary chat still works, which makes the error look inconsistent.
Thinking-model messages are metered on a weekly allowance on paid plans, so a research-heavy week can exhaust Thinking while your standard chat allowance is still fine.
OpenAI reduces limits when the service is under load, so the number you hit today may be lower than the number you hit last week or the figure quoted in a guide.
Every turn in a long conversation re-sends prior context, and attachments add processing weight. The same number of messages costs more in a bloated thread than in a fresh one.
When to try: First
If ChatGPT shows a specific time or countdown, that is authoritative for your account. Because the window is rolling, capacity returns gradually rather than all at once, so trying again in 20 to 30 minutes often gets you one or two messages even before the full window clears.
When to try: Immediately, to identify what is actually limited
Try a plain text message with no attachment and no image request. If that works, the block is on a specific feature (images, uploads, Thinking, Deep Research, voice). Switch to the part of the workflow that is not capped and come back to the rest later.
When to try: When only the reasoning model is blocked
Open the model picker and choose the fast or Instant option instead of Thinking. Lighter models have much larger allowances, and on Free tier ChatGPT already falls back to a mini model automatically once the standard cap is used up. Save Thinking for the questions that genuinely need it.
When to try: When you are deep in a long thread
Open a new conversation and paste in only the context you actually need. Long threads re-send their entire history on every turn, so a fresh chat with a short summary costs far less of your allowance per message.
When to try: To stretch a limited allowance
Instead of ten short back-and-forth messages, write one prompt that states the goal, the constraints, and the output format you want. Ask for numbered deliverables in a single response. Each request counts the same against your cap regardless of how much you asked for.
When to try: If you hit the cap far sooner than expected
Go to Settings, then your account or subscription page, and check the limits listed for your plan rather than relying on third-party figures. Numbers move, and Free, Go ($8/mo), Plus ($20/mo), Pro ($100/mo) and Pro ($200/mo) all differ substantially, with the Pro tiers effectively unlimited on the standard model.
When to try: If you hit the limit regularly
If you hit the cap most days, upgrading is the direct fix: Go raises Free caps cheaply, Plus raises them again with a weekly Thinking pool, and the Pro tiers give roughly 5x and 20x Plus limits. If you would rather not pay more, run the blocked task in Claude, Gemini, or Perplexity while your ChatGPT window resets.
When to try: Always, before reloading the page
Copy the message out of the composer into a notes app before reloading. Refreshing while capped can discard a long prompt, and rewriting it wastes more time than the wait itself.
Keep a scratch doc of long prompts so a cap or refresh never loses your work
Reserve the Thinking model for genuinely hard questions, use Instant for the rest
Start a new chat per task, long threads cost more allowance per message
Batch related questions into one structured prompt instead of many small ones
Contact OpenAI via help.openai.com only if you are blocked well below your plan's stated allowance, if a paid plan shows Free-tier limits after a successful payment, or if the limit never clears across a full reset window. Include your plan, the exact error text, timestamps, and roughly how many messages you sent, since limit disputes are resolved from account-side usage records.
Your account has used the messages allowed for that model within a rolling time window (commonly the trailing 3 to 5 hours depending on plan and model). It is a usage cap, not an outage. Access returns gradually as your oldest messages age out of the window.
Caps use rolling windows, not midnight resets, so there is no single reset time. Capacity comes back progressively as the messages you sent earlier fall outside the window, which is why you may get one message back before the full allowance returns. If the error names a specific time, trust that over any published figure.
Free runs on GPT-5.5 Instant with a small standard-model allowance before falling back to a lighter mini model. Go ($8/mo) and Plus ($20/mo) raise that substantially and add a separate weekly Thinking allowance. Pro at $100/mo gives roughly 5x Plus limits and Pro at $200/mo roughly 20x, both effectively unlimited on the standard model. Exact numbers shift with demand, so check your account's subscription page.
Three usual reasons: you were using the Thinking model, which is metered separately and more tightly; you were working in a very long thread, which re-sends its whole history every turn; or a per-feature cap (images, uploads, Deep Research, voice) ran out rather than your chat allowance. Caps also tighten during peak demand.
Usually yes. Switch to the faster Instant model, which has a much larger allowance, and avoid the specific capped feature. Free accounts are automatically moved to a lighter mini model instead of being blocked outright. Plain text chat frequently still works when image or upload limits are exhausted.