Claude Max 20x costs $200 a month. Max 5x costs $100. The names suggest four times the usage for twice the price. Anthropic only promises that multiple for the 5-hour session limit. It does not say how the weekly limit scales.

So we measured both plans on the same heavy Claude Code account, with hourly readings of the usage meter matched to the token logs.

The short answer: Max 20x gives about 2.07x the weekly limit and 3.57x the 5-hour limit of Max 5x, for twice the price.

Max 5xMax 20x20x vs 5x
Price per month$100$2002x
Weekly limit (API-equivalent)$1,248$2,5432.07x
5-hour limit (API-equivalent)$125$4453.57x
Usage per $1 of plan price, per month$54$55about equal

Per dollar, the two plans give you the same weekly usage. What 20x adds is room inside each 5-hour window. That decides how easily you actually get to 100% of the week.

How much more weekly usage does Max 20x give?

About 2.07x the weekly limit of Max 5x (95% interval: 1.78x to 2.31x). In API-equivalent dollars, a full week held about $1,248 on Max 5x and $2,543 on Max 20x.

“API-equivalent” means what the same tokens would cost on the Anthropic API at list price. It’s the easiest common unit for usage that mixes input, output and cached tokens. It’s a measure of usage, not money you pay. The 95% intervals: $1,110 to $1,407 for Max 5x, and $2,476 to $2,628 for Max 20x.

The dots in the chart at the top are real weeks: the highest weekly percentage the meter reached in that week, valued at the measured limit. The last six weeks on Max 5x peaked between 41% and 99%. Both full weeks on Max 20x reached 100%.

Why it isn’t 4x

Anthropic’s Max plan page defines the tiers per session: Max 20x “includes 20 times the Pro plan’s per-session usage allowance”, and Max 5x five times. Separately, “Max plans also have a weekly usage limit that applies across all models.” No multiple is given for that weekly limit, and in our measurement it does not follow the session multiple.

Why other numbers differ

The ratio moves with the period you measure. Classmethod measured about 3.5x in August 2026, with Max 5x at roughly $1,000 to $1,200 and Max 20x at $3,400 to $3,900 per week. That was during Anthropic’s weekly limits promotion, which raised Claude Code weekly limits by 50% from May 13 to September 13, 2026. Since September 14, Claude Code weekly limits are 25% above their pre-promotion level.

Our numbers come from after the promotion, on a different model and account. We can’t fully reconcile the two results. What they do agree on: the weekly ratio is not the 4x the plan names suggest, and nobody outside Anthropic knows the exact figure for your account.

The 5-hour limit is where Max 20x pulls ahead

The 5-hour limit scales much closer to the name: 3.57x (95% interval: 3.15x to 4.04x). A full 5-hour window held about $125 on Max 5x and $445 on Max 20x.

Bar chart of Max 20x divided by Max 5x. Price: 2x. Weekly limit: 2.07x, interval 1.78 to 2.31. 5-hour limit: 3.57x, interval 3.15 to 4.04. A dashed line marks the 4x the plan names suggest.
For twice the price, Max 20x gives about twice the week and three and a half times the 5-hour window.

That difference matters more than it looks, because the two limits interact. On Max 5x, one full 5-hour window is 10% of the week. To use the whole week you need ten maxed-out windows. On Max 20x, one window is 17.5% of the week, so about six full windows use it up.

Two bars each representing a full week. The Max 5x bar is split into 10 segments, one per maxed 5-hour window. The Max 20x bar is split into 5.7 segments.
A week measured in maxed-out 5-hour windows: 10 on Max 5x, 5.7 on Max 20x.

A working day holds two or three 5-hour windows. On Max 5x, using the full week therefore means three to four days of hitting the 5-hour limit, window after window. On Max 20x, two heavy days can carry most of the week.

What that looks like in practice

The chart below shows every 5-hour window the account used, by how full it got. The last two weeks on Max 5x come first, then the first two weeks on Max 20x.

Dot plot of 5-hour windows by percentage used. On Max 5x, 14 windows hit the limit in 13 days. On Max 20x, most windows used 15 to 35 percent and 1 window hit the limit in 14 days.
Max 5x: 14 windows hit the limit in 13 days. Max 20x: 1 window in 14 days.
  • Max 5x: 14 windows hit the 5-hour limit in 13 days. A used window typically reached 74% of its limit. The busiest day used 33% of the week.
  • Max 20x: 1 window hit the limit in 14 days. A used window typically reached 32%. The busiest day used 62% of the week.

The usage pattern was not identical: the account was used more heavily after the upgrade. The direction is still the point. On Max 5x the 5-hour wall cuts heavy days short, so the weekly limit is hard to reach. On Max 20x the wall rarely comes, and the weekly limit becomes the only one that matters.

Is Max 20x worth it?

Per dollar, the plans are equal: about $54 of usage per $1 on Max 5x and $55 on Max 20x, per month at full use (the weekly limit times 4.33 weeks, divided by the price). The ratio of the two is 1.04 (95% interval: 0.89 to 1.15). The extra $100 for Max 20x buys about $1,295 more usage per week, which is $56 per extra dollar.

So the question is not value per dollar. It is whether you can use the weekly limit you pay for.

Your typical weekMax 5xMax 20x
Light use, rarely hits the 5-hour limitSame value, half the priceMore room than you use
Hits the 5-hour limit most days, week ends below ~90%Blocked during the dayWorth it: the 5-hour wall mostly disappears
Needs more than ~$1,250 of usage a weekRuns outWorth it
Needs more than ~$2,500 a weekRuns out earlyRuns out too

To see where you are, open the usage page in Claude settings or run /usage in Claude Code. Watch two things: how often you hit the 5-hour limit, and how full your week is when it resets.

What 1% of the meter is worth

Meter1% of the limit equals
Max 5x, 5-hour$1.25
Max 5x, weekly$12.50
Max 20x, 5-hour$4.45
Max 20x, weekly$25.40

At this account’s typical pace of $25 to $33 of usage per active hour, the Max 5x 5-hour meter moves 20 to 26% per hour. On Max 20x it moves 6 to 7%. The Max 20x weekly meter moves about 1% per active hour, so it fills after roughly 80 to 100 active hours.

What happens to your usage when you upgrade mid-week

The account upgraded from Max 5x to Max 20x in the middle of a week, with the weekly meter at 77%. The next reading, about an hour later, showed 4%. The weekly counter started over at the upgrade, and the reset day stayed the same. Billing was prorated for the rest of the month.

That is one observation, not a documented rule. Anthropic’s page says the weekly limit “resets at a fixed time each week that is assigned to your account” and that “you receive your full weekly allowance each cycle”. It does not describe upgrades.

How we measured

  • Meter readings: CodexBar records the Claude usage meter (5-hour and weekly percentage) about once an hour and keeps the history locally.
  • Usage: every request in Claude Code’s local logs, including scripted claude -p runs, priced at Anthropic API list rates for Opus 5.5: $4 per million input tokens, $20 output, $0.20 cache reads, $5 or $8 cache writes (5-minute or 1-hour cache).
  • Calibration: for each pair of consecutive meter readings, we took the change in percentage and the usage in between. The limit is the total usage divided by the total percentage gained. The 95% intervals come from resampling whole days (bootstrap, 4,000 draws).
  • Sample: Max 5x from September 22 to 27, 2026 (131 weekly and 100 five-hour intervals). Max 20x from September 27 to October 11 (332 weekly and 273 five-hour intervals). An earlier Max 5x stretch, September 14 to 22, gives similar values ($1,211 per week, $131 per 5 hours) with wider intervals.

The check that matters: does usage in dollars actually explain the meter? On Max 20x it does, closely. Usage explains 97% of the variation in the weekly meter per 6-hour block.

Scatter plot of weekly meter percentage gained against API-equivalent usage per 6-hour block on Max 20x. The points follow a straight line of 1% per $25.43, with an R squared of 0.97 across 57 blocks.
Max 20x weekly meter against usage per 6-hour block: 1% per $25.43, R² = 0.97.

Limits of this measurement:

  • It is one account, used almost entirely with Opus 5.5 in Claude Code.
  • The Max 5x sample is short (six days).
  • The meter shows whole percentages.
  • Anthropic’s internal weights are not public. Dollar value is a proxy that fits well here, but other models or usage mixes may weigh differently.
  • Anthropic can change limits at any time. These numbers describe September and October 2026.

FAQ

Is Claude Max 20x four times Max 5x?

Only per session, and only roughly. In our measurement Max 20x gave 3.57x the 5-hour limit and 2.07x the weekly limit of Max 5x, for twice the price. Anthropic publishes the multiple for the session limit only.

How much more weekly usage does Max 20x give than Max 5x?

About twice as much: 2.07x, with a 95% interval of 1.78x to 2.31x, measured on one account in September and October 2026. During Anthropic’s May to September 2026 promotion, one other measurement found about 3.5x.

Is Max 20x worth the extra $100?

It gives the same usage per dollar as Max 5x, about $55 of API-equivalent usage per $1. It is worth it if you hit the 5-hour limit most days or need more than about $1,250 of usage a week. If neither is true, Max 5x gives the same value for half the price.

Does upgrading reset the weekly limit?

In our one observation, yes: the weekly meter went from 77% to 4% right after the upgrade, and the reset day stayed the same. Anthropic does not document this, so treat it as likely rather than guaranteed.