Claude Max 5x vs 20x: which tier you actually need

The difference is not quality — it is how many times you can hit the wall before lunch. A practical way to size your plan.

Both Max tiers give you the same models, the same Claude Code integration and the same context window. Nothing is unlocked at 20x that is locked at 5x. The only difference is the volume of usage before you are throttled, and that makes the choice a pure arithmetic question about your working day.

Put the prices on the table first

Here, Claude Pro is 105 CNY (about 14 euro), Max 5x is 523 CNY (about 70 euro) and Max 20x is 1047 CNY (about 140 euro). The official equivalents, 134, 668 and 1336 CNY, sit next to them on each product page. Every plan is one month and does not renew by itself, so the gap you are deciding over is 524 CNY for a single month of headroom, not a permanent contract.

What the multipliers actually mean

Anthropic does not publish an exact token number for either tier. The multipliers are relative to Pro, and the real limit is enforced through two overlapping windows:

• a short rolling window — roughly a five-hour block, after which you wait for the timer to move;
• a weekly ceiling — which catches people who never hit the short window but burn steadily all week.

That structure matters more than the headline multiplier. Two developers can both be on Max 5x and have completely different experiences: one hits the five-hour wall every afternoon, the other never touches the weekly ceiling. Which wall you are likely to meet decides the tier more than your total hours do.

Who is fine on 5x

Choose 5x if most of your usage is interactive: you ask, you read, you edit, you ask again. A typical day of chat plus moderate Claude Code use in a single repository rarely saturates five times the Pro allowance. If you have never once seen a rate-limit notice on Pro, 5x is already more than you need.

Who needs 20x

The 20x tier starts to pay off when work runs without you. Concretely:

• agentic runs that iterate over a large repository for tens of minutes;
• parallel sessions — one for the backend, one for the frontend, one for tests;
• long refactors where stopping mid-task costs you more than the subscription;
• any workflow where a rate-limit notice actually breaks something.

If you are interrupted more than once or twice a day, the upgrade is cheaper than the lost focus. If you are interrupted once a week, it is not.

The honest math

What the multiplier does not buy

20x does not give you early access to new models, it does not raise the context window, and it does not remove the per-request caps that apply to very long generations. People upgrade expecting a different product and are disappointed. You are buying headroom, nothing else.

A cheaper way to decide

Start on 5x for one billing period and keep a note of every rate-limit hit — time of day, what you were running, whether it actually blocked you. After thirty days you will have a real number instead of a feeling. The people who ask us for 20x before they have ever hit a limit on 5x usually want peace of mind rather than capacity; the ones who genuinely need it can describe, without thinking, the two or three moments in the week where the run stopped.

If you buy the wrong one

An order here is one month, and a refund is computed as the order amount times the unused days divided by the total days, with a replacement offered first. That means upgrading mid-month is a cheap experiment: the unused part of the 5x order has a stated value, and the difference is what you pay for the rest of the period.

If you already know you need the top tier, the Max 20x listing shows the current price and delivery window. For how the two windows behave in practice, see how Claude Code rate limits work.

Common questions

Back to plans →