Claude Pro Get How Many Tokens? 5-Hour Limit

Claude Pro include token how many?

Claude Pro no get token allowance, and as of October 2026 Anthropic no publish one. Claude subscription no be token balance wey you dey use up. Anthropic dey measure usage with tokens, but Pro limit am with session limit wey dey reset every five hours, plus weekly limit wey apply across all models. Dem no publish either limit as token figure. How long one session fit last depend on the model wey you choose, how long your conversation don grow, and how much text each turn carry come.

Na this be the answer wey this question rarely get, because people expect plan make e work like API key. API key dey spend balance wey dem count with tokens, so you fit calculate the invoice by yourself. Allowance-based plan like Pro, Max or Team dey sell access to cap wey Anthropic set and adjust. Anthropic no state this cap as token number, and the Pro help page only talk say the number of messages wey you fit send go vary. So no fixed token number dey to quote. Anybody wey give you one don make am up.

So ask different question. Wetin dey make my messages expensive? That one get real answer, and you fit act on am today.

Wetin make subscription no fit quote token number

The API dey bill every request by tokens. Input tokens cover everything wey you send, while output tokens cover everything wey the model write back. The two get different rates, na why input and output tokens cost different amounts before you estimate anything.

The consumer plans no get balance wey dem fit spend for the chat usage wey dem include. Apart from verified Ultra accounts, Anthropic dey cap your usage inside session wey dey reset every five hours. The allowance-based paid plans (Pro, Max, Ultra, Team and older seat-based Enterprise contracts) still get another longer limit wey dem measure across one week. Usage based Enterprise plans dey work different: Anthropic talk say dem no get per-seat usage limits and no included token allowance. Every token dey bill at API rates. Instead, administrators fit set spend limits for the organization and for individual users. Message no get fixed size, so the same message count fit mean very different amount of work for two different people. Anthropic's Pro help page no give any message count at all. E talk say the number of messages wey you fit send dey depend on message length, the length of files wey you attach, the length of your current conversation, and the model or feature wey you use. Na this warning explain the whole matter, but dem almost never explain the mechanism behind am.

Why turn 20 for chat dey cost far pass turn 1

The model no dey keep memory between requests. E dey stateless, so every request must send the conversation context again: the system prompt, earlier messages and replies, plus any attached content wey still dey for context. Claude Code dey compact long conversations into summary, so that context no always contain the complete original transcript. Your new question fit be twenty words. But the request wey carry am still include the existing context.

So, input token count for one turn dey grow as chat length dey increase, and for this example e dey grow in straight line. The rows below show the calculation from one set of rough assumptions: system prompt near 1,000 tokens, your messages near 1,000 tokens each, and replies near 1,200 tokens each. These figures dey show how the mechanism work. Dem no be measurement of your account.

[
  {
    "label": "Turn 1",
    "sent_this_turn": "2,000",
    "cumulative_input": "2,000"
  },
  {
    "label": "Turn 5",
    "sent_this_turn": "10,800",
    "cumulative_input": "32,000"
  },
  {
    "label": "Turn 10",
    "sent_this_turn": "21,800",
    "cumulative_input": "119,000"
  },
  {
    "label": "Turn 20",
    "sent_this_turn": "43,800",
    "cumulative_input": "458,000"
  }
]

Turn one sends 2,000 input tokens. By turn 20, the same short question sends 43,800, wey pass twenty times as much, even though na the same amount of typing you do. Across the full conversation, you don send 458,000 input tokens, and most of am na the same text wey dey send again and again.

Twenty separate one-turn chats for send twenty times the first row above, and nothing else. Na the same questions, but only small part of the load. Attachments make the difference bigger when their contents remain for context, because that content counts as input again for later turns.

Na why two people on the same plan fit report very different mileage. One person keep one thread open throughout the week and reach the cap by Wednesday. The other person open fresh chat for every task and rarely see any limit. Neither person dey do anything wrong. Their token spending differ by one order of magnitude because their usage habits differ.

For the API, prompt caching fit reduce the repeat cost. E stores the unchanged beginning of a request, so later calls get reduced input rate for that part. Claude Code dey do this automatically for you, both with subscription and API key, and e let you disable am or choose how long the cache go remain. Caching make the repeat cheaper, but e no remove am: every turn still carry the current context, so conversation length na the control wey you really get. The same mechanism affect coding sessions, where file contents and loaded tool definitions add to the context. So keeping coding session context small help preserve your limits, and where Claude Code session tokens really dey go worth reading before you blame the plan.

Wetín each plan tier dey give you, compared to others

No subscription tier dey publish one fixed allowance. Wetin dem publish na relative limit, and this comparison enough to help you choose. Free dey for bottom, and Anthropic talk say number of messages wey free user fit send dey change based on demand. Pro dey above am. Max cost $100 every month for web, and e include five times Pro usage every five hours and every week. The former Max 20x plan don become Ultra, at $200 every month for web, with ten times Pro weekly usage and no five-hour limit for verified accounts. Team dey priced per seat, and the limits apply to each member, no be the whole team together. Usage based Enterprise plans na the exception for this whole page: dem charge seat fee wey cover platform access only, then bill every token separately with standard API rates, plus 1.1x multiplier for US-only inference on Opus 4.6, Sonnet 4.6 and later models. Administrators fit set spend limits, and self-serve Enterprise usage also stop for the whole organization when prepaid credits finish, until person wey get billing permission buy more. Every consumer cap belong to an account, no be one person, so two people wey dey share one Pro login go drain one window between dem, and the consumer terms still forbid the arrangement.

Treat those multipliers as statement of intent, no be contract. Anthropic dey adjust the caps, and its help pages reserve the right to limit usage in other ways, like weekly and monthly caps or limits on particular models and features, whenever e choose. Na another reason why no honest page fit give you one token number. For the money side of this same comparison, wetin Pro cost and where e dey limit you and the difference between Max and Ultra cover am, while wetin the free plan really allow cover the lowest level. If na from Naija you dey pay, wetin Pro go cost you for naira compare App Store price with claude.ai price for naira card, and e show you how to calculate am with your own bank rate.

Model choice na second multiplier wey dey on top of the tier. The same question go use more of your window on the biggest model than on the smallest one, because heavier model cost more tokens to run. If you move routine work from Opus to Sonnet or Haiku, e often give you more working time than upgrade, so matching the model with the job na the first thing to try when you dey run out. If that routine work later move go the API, first check wetin Haiku 5.5 cost per million tokens, because prompts wey pass 100K tokens dey cost five times the normal rate. So the long-thread habit wey dey drain subscription window go cost more there too.

How to see your own usage instead of guessing

Your own account na the only source wey get authority, and e dey just two clicks away. For claude.ai, open Settings, then Usage. E go show wetin you don use and when your limits go reset. Check am once when replies start to dey get refused, and once after long session. You go learn your own pattern faster than any published figure fit teach you. The app still dey warn you before that point: a banner wey talk say you don use 90% of your session limit mean say na only one-tenth of the current session remain. Na that time you suppose open fresh chat or move to lighter model, instead of pushing long thread further.

For Claude Code, two slash commands dey do the same work from terminal. /usage dey report your plan usage limits, and for Pro, Max, Team and Enterprise plans e dey add breakdown of wetin count against dem. /context dey show wetin dey fill the context window now, with breakdown for each item and suggestions about wetin to trim. When /context show say old conversation dey take most of the space, /clear dey start new conversation with empty context, and the cost per turn go drop back near the first row of the chart above.

For the API, you fit ask for count before you send anything:

curl https://api.anthropic.com/v1/messages/count_tokens \
  --header "x-api-key: $ANTHROPIC_API_KEY" \
  --header "anthropic-version: 2023-06-01" \
  --header "content-type: application/json" \
  --data '{
    "model": "claude-sonnet-5",
    "messages": [{"role": "user", "content": "Summarise the attached report."}]
  }'

Healthy response na one field, shown here with illustrative count:

{"input_tokens": 14}

Anthropic dey label that number as estimate. The count no talk anything about output wey model never write yet. Every completed request dey report the actual accounting when e finish:

{"usage": {"input_tokens": 21430, "output_tokens": 512}}

Send one short question as fresh request. Then send that same question at the end of long thread, and compare the total input tokens. When caching dey used, add cache_read_input_tokens and cache_creation_input_tokens to input_tokens. The difference show how much extra context the long thread carry. Cached content dey count less against subscription limits when dem reuse am.

Wetin to do when limit land for middle of task

1. Wait for the window. The app go show the reset time, and the wait often shorter pass the workaround. Na the correct move be this when nothing dey due. If your account get limit reset wey you fit use, Anthropic's Pro help page talk say you fit use am restore your five hour session limit or weekly limit to full, and how limit resets dey work explain when e make sense to do am. If you no wan wait, Pro subscribers fit enable usage credits to continue work beyond the limits wey the plan include.
2. Move go lighter model. Smaller model cheaper per turn for every plan. For Max, Anthropic's Fable guidance allow you spend up to half of your weekly limit on Fable models without extra cost, and e dey come from the same weekly limit wey every other model use. So after you don use that share, you fit switch to another model or continue to use Fable with usage credits. For Pro, Fable no dey inside the plan limits at all, and e dey run on usage credits from the first message.
3. Start fresh chat wey carry only wetin the next step need. Paste the conclusion, no be the transcript. Cost per turn go drop immediately, and answers normally improve because the model no longer dey weigh thousands of tokens from discussion wey don settle.
4. Move the job go API, where tokens dey count per request and dem bill per token. You no get five hour session to wait out, but API get im own windows. Rate limits dey measure per minute for requests, input tokens, and output tokens. If request pass any of dem, e go return 429 error, usually with retry-after header wey talk how long to wait. The Start, Build and Scale usage tiers also get monthly spend cap of $500, $1,000 and $200,000. Once you reach the cap, API usage go pause until 00:00 UTC for the first day of next month, unless you request higher limit before then. You fit estimate call input before you send am and read the exact usage after.

Which option fit depend on whether na clock or workload dey cause your problem. One-time rush na timing problem. Wall wey you dey hit every Wednesday na workload problem, and e need API or bigger plan instead of better timing. the complete checklist for limit wey land for middle of task go guide you through recovery step by step, and how rolling windows dey open and reset explain why the clock dey behave the way e dey behave. If na API you dey plan use, price am first: API compared with subscription for the same workload and wetin one million tokens really cost give you the calculation before you move.

FAQ

Pro include token kam?

Anthropic no publish token allowance for Pro, and as of October 2026 no such number dey available to quote. Pro usage dey measured in tokens against session limit wey reset every five hours and weekly limit across all models. How many messages fit enter dey depend on the model, your attachments, and the length of your conversation. Any page wey state specific monthly token figure for Pro dey guess. Open Settings, then Usage for claude.ai to see your own consumption and reset time.

Starting new chat go reset my usage?

No. Usage dey counted for your account across the window, so new chat no go return wetin you don already spend. Wetin e change na the price of every turn from that point. New chat no get transcript to send again, so e start for the low end of the chart above instead of continuing to increase. Opening fresh chat when topic change na the cheapest habit wey dey available for any plan.

Why I hit the limit faster today than yesterday?

Most times na because today's work dey inside one long thread. Earlier messages and attached content wey still dey for context count again as input, so input per message dey grow until context trim or compact am. The other common cause na the model: heavier models use the allowance faster for each turn, so afternoon on the largest model fit finish earlier than morning on smaller one.

I suppose move to API to get real token counting?

Move if you need the numbers or predictable usage. API report input and output tokens for every call. E expose free token counting endpoint wey you fit run before sending, and e no get five-hour session wey you must wait out. E still get limits: per-minute rate limits wey return 429 error, usually with wait time, when you exceed dem. For Start, Build and Scale tiers, e also get monthly spend cap wey pause usage until the first day of the next month after you reach am. E bill per token, so heavy use cost more than flat subscription, while light use cost much less. Price your real workload against your current plan before you switch, because the answer depend on how many tokens you send.