back to blog
EXPLAINERNovember 16, 2026·7 min read

AI limits: message caps and rate limits in ChatGPT, Claude and Gemini

How AI usage limits work: message caps, rate limits and context windows in ChatGPT, Claude and Gemini, what 'unlimited' means, and how to tell which you hit.

By Unrestricted team · Editorial standards

Every major hosted AI chat has usage limits, and they come in kinds that are easy to confuse: message caps (how much you can use per period), rate limits (how fast you can send), file and context-window limits (how much one chat can hold), and content rules (what it will answer). Plans sold as unlimited still keep some of them.

ChatGPT, Claude and Gemini set their limits by plan and publish them in their own help and pricing pages, and Anthropic and Google both say theirs can change. [1] [2]This guide covers usage limits. Content rules, meaning what a chatbot will and won't answer at all, are a separate question, covered in AI without restrictions: what it means, and its limits. Below: each kind of limit, what the three assistants' pages said when we checked them in November 2026, what "unlimited" means in their wording, and how to tell from the message which limit you hit.

The kinds of limits: messages, rate, files and context window

Message caps.A message cap limits how much you can use an assistant over a window of time. Claude's paid plans (except verified Ultra accounts) have a session limit that resets in full five hours after your first message, plus weekly limits, and usage on claude.ai, Claude Code and Claude Desktop counts toward the same limit. [1] Gemini Apps say your limit refreshes every five hours until you reach your weekly limit. [2]A cap isn't always a fixed number of messages: Anthropic lists message length, file attachment size and the length of the current conversation among the things that use it up [3], and on Claude's free plan the number of messages you can send varies with demand. [4]

Rate limits.A rate limit caps how much you can send in a short, fixed window. OpenAI's API, for example, measures its rate limits in requests or tokens per minute or per day, images per minute and, for some streaming audio models, audio minutes per minute. [5] OpenAI says they protect against abuse, keep access fair and manage the load on its infrastructure. [5]

File limits. Uploads have caps of their own. Claude takes up to 20 files per chat, at up to 500MB each. [6] Gemini Apps take up to 10 files in one prompt, subject to availability, with videos up to 2GB and other files up to 100MB, and add rolling limits on how much you can upload and analyze in a period of time. [7]ChatGPT's pricing page lists messages with uploads as limited on the Free plan. [8]

Context window. The context window is how much a model can work with in a single chat, measured in tokens. [1]Part of it goes to the reply and to instructions the service adds, so the room for your own input is smaller than the full window: Anthropic reserves part of Claude's window for its response [1], and OpenAI says system instructions, memories and internal processing share ChatGPT's window. [8] On ChatGPT and Gemini, the window grows with the plan:

Context windows as each company states them, checked November 10, 2026. ChatGPT's page gives its figures without a unit; Claude's and Gemini's give tokens.
AssistantWithout a paid planPaid plansSource
ChatGPTGPT Instant: 27K, an input maximum of about 12 pages of text. GPT Reasoning: varies.Go and Plus: 54K (Instant), about 40 pages of input, and 256K (Reasoning). Pro: 128K (Instant), about 250 pages of input, and 400K (Reasoning).Pricing page [8]
ClaudeNot stated on the pages checked.The newest models support up to 1M tokens; others 500K or 200K.Help Center [1] [9]
Gemini32K tokens.Google AI Plus: 128K. AI Pro and AI Ultra: 1 million.Help Center [2]

Content rules.These decide what an assistant will answer at all, which is a different kind of limit from the four above. A refusal is a content rule, and waiting for a reset won't change it.

Message limits by assistant

Here is what each company's own pages said about message limits, plan by plan, on the date in the last column. Limits change, so check the linked page before you rely on a row.

Checked November 10, 2026. Each assistant row links its source; the Unrestricted row is written from our own plan rules.
AssistantFree planPaid plansSource, checked
ChatGPTEveryday text chats and chat history listed as unlimited, with text chats subject to abuse guardrails. Messages with uploads, image creation, voice chats and deep research are listed as limited.Everyday text chats listed as unlimited on Go, Plus and Pro too. Go: deep research listed as limited. Plus: expanded messages and uploads, with a note that limits apply. Pro: a choice of three usage tiers; image creation, voice and several named models are also listed as unlimited, subject to abuse guardrails. A footnote on every plan's unlimited entries says usage must be reasonable and comply with OpenAI's policies.Pricing page, Nov 10, 2026 [8]
ClaudeA session-based limit that resets every five hours. The number of messages varies with demand.Pro: a five-hour limit and a weekly limit across all models [10]. Max: five times Pro's usage every five hours and every week [11]. Ultra: 10 times Pro's usage every week, and no five-hour limit for verified accounts [12].Help Center, Nov 10, 2026 [4]
GeminiStandard limits. The page gives no number.Google AI Plus: twice the standard limits. AI Pro: four times. AI Ultra: five or 20 times AI Pro's, depending on the subscription.Help Center, Nov 10, 2026 [2]
UnrestrictedA free allowance to try it. On the web it is one-time and doesn't reset, so unlike Claude's and Gemini's free limits, it doesn't refresh over time.Pro. On the web, Pro removes the message quota; per-minute rate limits still apply. iPhone limits are set separately.Our own plan rules, Nov 2026

None of the three pages gives a fixed message count for any plan in the table. Paid usage is described as more than another plan, a multiple of it, or unlimited with conditions. If you need a precise budget, check inside the app: Anthropic says users on Pro, Max, Ultra, Team and seat-based Enterprise plans can see progress toward their limits under Settings, then Usage [3], and Gemini shows your usage under Settings, then Usage Limits. [2]

What "unlimited" plans actually mean

Where these pages use the word, it comes with conditions. OpenAI's pricing page lists everyday text chats and chat history as unlimited on every consumer plan, Free included, and the Free and Go plans carry a footnote that reads "Unlimited text chats subject to abuse guardrails." Every plan's unlimited entries share another footnote: usage must be reasonable and comply with OpenAI's policies, and usage refers to chat limits. On Pro, the unlimited entries also include image creation, voice and several named models, again subject to abuse guardrails, while the Free plan still lists uploads, image creation, voice chats, deep research, and memory and context as limited. [8]So on ChatGPT, "unlimited" covers specific features, under conditions, and the context window in the table above still applies.

Anthropic's plan pages don't call a plan unlimited. They describe five-hour and weekly limits, say Anthropic may limit usage in other ways at its discretion and add limits to manage capacity and keep access fair, and say it doesn't guarantee uninterrupted service. [10] [11] [12] Its usage article does take up the question of unlimited usage, and its answer for paid plans is to buy usage credits. [1] Google says Gemini Apps limits may change without notice, including because of capacity, and that it may at times cap the number of prompts and conversations you can have. [2]

Our own Pro is no exception: on the web it removes the message quota, and per-minute rate limits still apply. When a plan says unlimited, ask three questions:

  1. 01

    Unlimited what?

    Text chat alone, or every feature? A plan can drop the cap on chat while uploads, voice or research tools stay limited.

  2. 02

    Under what conditions?

    Look for the footnote: abuse guardrails, reasonable use, capacity, or the right to change limits without notice.

  3. 03

    What still applies?

    Rate limits, file limits and the context window can all stay in place on a plan with no message quota.

Which limit did you hit? Reading the message

The wording of the notice can tell you which kind of limit you ran into, and so whether waiting, a new chat or a bigger plan will fix it.

It names a time window or a reset time: a message cap. Claude's notices name the five-hour limit or the weekly limit and when it resets, and Claude warns you when you are getting close; if usage credits are turned on, the notice says it is continuing with credits instead. [13] Gemini Apps notify you when you are close to your limit and again when you reach it, with the time it will refresh, and Google AI subscribers can continue the conversation with Flash-Lite. [2] Waiting for the reset or moving to a bigger plan fixes it.

It says to slow down or try again soon: a rate limit or a capacity problem.Developers using OpenAI's API get a 429 error when a request goes over a rate limit, and the response can say how long to wait before retrying. [5] Claude has a separate message for capacity constraints, which says Claude is unable to respond and to try again soon; Anthropic says a capacity constraint is not an outage, and its advice is to try again in a few minutes. [13]

It says your message or the chat is too long: the context window.Claude's version says your message will exceed the length limit for this chat; Anthropic suggests splitting or condensing the content, attaching fewer or smaller files, or starting a new conversation. [13] On Gemini there may be no message at all: Google says going past the window can produce answers that miss some of what you provided. [2] A bigger plan can raise the window on ChatGPT and Gemini [8] [2], but every plan in the context-window table above has one.

It's about a file: an upload limit. Too many files, a file that is too large, or, on Gemini, the rolling limit on how much you can upload in a period of time. [7]

It declines the request itself: a content rule. That is a refusal, not a usage limit, and no reset changes it. How refusals are produced, from training to filters added on top, is explained in how AI content filters work.

Unrestricted's limits

On Unrestricted, the web's free allowance is one-time and doesn't reset; after it, Pro removes the message quota on the web, and per-minute rate limits still apply. iPhone limits are set separately, web and iPhone subscriptions are billed separately, and the Unrestricted FAQ covers accounts, plans and what the model provider still refuses.

If refusals are your problem rather than caps, start with what AI without restrictions can and can't mean. If cost is, our guide to free uncensored AI compares the free routes, from a model on your own computer to a hosted chat's free allowance, and the catch with each.

Frequently asked

  • Why does ChatGPT say I've hit my limit?

    It can be a feature cap rather than a chat cap. ChatGPT's pricing page lists everyday text chats as unlimited on every plan, with a footnote that usage must be reasonable, while messages with uploads, image creation, voice chats and deep research are listed as limited on the Free plan. If the message is about length, it points to the context window, and starting a new chat can help.

  • Is 'unlimited' really unlimited?

    Not on the plan pages we read. ChatGPT lists everyday text chats as unlimited on every plan, with a footnote that usage must be reasonable and comply with OpenAI's policies, and its Free and Go plans add that text chats are subject to abuse guardrails. Anthropic and Google say they may cap or adjust usage to manage capacity. On Unrestricted, Pro removes the web message quota, but per-minute rate limits still apply.

  • What is a rate limit?

    A rate limit caps how much you can send in a short, fixed window, such as requests or tokens per minute or per day. OpenAI says its API rate limits protect against abuse, keep access fair and manage load, and that a request over the limit gets a 429 error, sometimes with a note on how long to wait. Waiting and sending again is the fix, and paying for a bigger plan doesn't always remove it.

  • What is a context window?

    A context window is how much a model can work with in a single chat, measured in tokens. Part of it goes to the reply and to instructions the service adds, so the room for your own input is smaller than the full window. Paid plans can get a bigger one: Gemini Apps list 32K tokens without a plan and up to 1 million on AI Pro and AI Ultra. When a chat outgrows it, start a new chat or split long input.

Sources

Checked November 2026.

  1. How do usage and length limits work?, Anthropic (Claude Help Center)
  2. Gemini Apps limits & feature availability, Google (Gemini Apps Help)
  3. Usage limit best practices, Anthropic (Claude Help Center)
  4. Get started with Claude, Anthropic (Claude Help Center)
  5. Rate limits, OpenAI (API documentation)
  6. Upload files to Claude, Anthropic (Claude Help Center)
  7. Upload & analyze files in Gemini Apps, Google (Gemini Apps Help)
  8. Plans and Pricing, OpenAI
  9. How large is the context window on paid Claude plans?, Anthropic (Claude Help Center)
  10. What is the Pro plan?, Anthropic (Claude Help Center)
  11. What is the Max plan?, Anthropic (Claude Help Center)
  12. What is the Ultra plan?, Anthropic (Claude Help Center)
  13. Troubleshoot Claude error messages, Anthropic (Claude Help Center)