Anthropic sells you a limit and never says how big it is. Pro, Max and Team do not come with a number of messages or tokens. They come with a usage allowance that refills on a timer, and Anthropic describes each plan’s allowance as a multiple of another plan.
That design makes one question hard to answer: how much can I actually do? I read Anthropic’s help articles on the limits on 09/28/2026, and every fact below comes from them.
The three rules
1. One pool covers every Claude app
Anthropic’s help article “How do usage and length limits work?” states that your usage of “all different Claude product surfaces (claude.ai, Claude Code, Claude Desktop) counts towards the same usage limit.”
In practice, a two-hour Claude Code session in the morning leaves less for chat in the afternoon. There is no separate coding allowance.
2. The session limit resets every five hours
The Max plan article states that your “session-based usage limit will reset every five hours.”
3. A weekly limit sits on top
The same article describes a weekly usage limit across all models. It resets at a fixed time each week that is assigned to each account. You can stay under every five-hour limit and still hit the weekly one if you work long days all week.
Anthropic also reserves the right to apply “weekly and monthly caps or model and feature usage, at our discretion.” The published limits are not a guarantee that nothing tighter applies.
How much each plan gets
Anthropic publishes multiples, not numbers. From claude.com/pricing, read 09/28/2026:
| Plan | Price | Usage, as Anthropic states it |
|---|---|---|
| Free | $0 | Baseline |
| Pro | $20 a month | At least 5x Free per five-hour session |
| Max 5x | $100 a month | 5x Pro per five-hour session |
| Max 20x | $200 a month | 20x Pro per five-hour session |
| Team Standard | $25 a seat monthly | More usage than Pro |
| Team Premium | $125 a seat monthly | 5x Team Standard seats |
Because the baseline itself is never stated in tokens, the table tells you how plans compare to each other, not how much work any of them buys. I go through what that means for the upgrade decision in Claude Pro vs Max. The full price list, including the API, is in Claude pricing.
What uses the limit up
The usage-limits article names four factors:
- Conversation length and complexity. This is the big one, and the next section explains why.
- Features. Each tool or extended feature you turn on adds work to a message.
- The model. Anthropic lists the model as a factor. I think the largest models run the limit down fastest, though Anthropic publishes no ratio.
- The effort level. Higher effort means more thinking tokens per answer.
Why long conversations burn the limit
A language model has no memory between messages. Every time you send a message, the app sends the whole conversation so far along with it. Message one sends a few hundred tokens. Message forty sends everything from messages one to thirty-nine, plus your new one.
So the cost of a conversation does not grow in a straight line. Each message costs more than the one before it. A long thread with a large pasted document can use more of your limit in one reply than a dozen short fresh chats.
You can see this directly. Paste the text of a long conversation into the Claude token counter and it shows how many tokens that single re-send carries, and what it would cost on the API. The same number is roughly what each new message in that thread has to carry.
What to do when you hit the limit
The usage-limits article lists three options on paid plans:
- Wait for the reset. The session limit refills five hours after the window started.
- Upgrade. Pro to Max 5x is five times the allowance. Max 5x to Max 20x is four times more again.
- Buy usage credits. Usage credits let you keep working past the limit and pay for the extra.
Claude Code adds a fourth option. Its help article, updated 08/19/2026, says you can switch to API credits from the Claude Console, billed at standard API rates. Claude Code does this only with your explicit consent, so a plan never silently turns into an API bill.
To check where you stand in Claude Code before you hit the wall, run /status. The Claude Code specific version of this problem, with ten habits that make a plan last, is in Claude Code usage limits.
How to make the limit last
- Start a new conversation for each new task. This resets the size of what each message re-sends.
- Do not paste what you can summarize. A 40-page document pasted into a thread is carried by every message after it.
- Match the model to the task. Use a smaller model for rewording, lookups and short questions.
- Lower the effort level on routine work. Save high effort for problems that need it.
- Batch your questions. One message with three questions costs less than three messages that each re-send the thread.
A limit you cannot see is a limit you have to measure. The person who checks /status before the wall is way ahead of the one who finds it.
Tracking changes to the limits
Anthropic edits these articles and its prices. Two of the help articles I cite here were updated in the last six weeks. If you plan work around a limit, a quiet change can cost you a day.
I am building Claude Cost Watch, a short email that flags when a Claude price or limit changes. It is a waitlist today, and nothing is charged.
Sources
All read on 09/28/2026.
- Anthropic Help Center, “How do usage and length limits work?”
- Anthropic Help Center, “What is the Max plan?”
- Anthropic Help Center, “Use Claude Code with your Pro or Max plan” (updated 08/19/2026)
- Plan prices and usage multiples: claude.com/pricing
No link on this page is an affiliate link.