Press Esc to close

Anthropic is reportedly A/B testing Claude Max usage limits for some subscribers

Anthropic is reportedly A/B testing Claude Max usage limits for some subscribers
Two glowing usage gauges side by side, one orange and one blue, with needles at slightly different positions

Image: TechcityAuthority

Anthropic is reportedly running an A/B test on usage limits for Claude Max, according to research firm SemiAnalysis. In a post on X, the firm said it detected a test that cuts monthly API-equivalent value by about 20% for some accounts while raising five-hour limits by about 20%. It added that the change only affects some users.

SemiAnalysis backed the claim with a chart. It measured two Max 20x accounts on the same day with the same model, Opus 5.5. One account had a five-hour window of about 2,076 million agentic tokens, against 1,740 million on the other, a rise of 19%. Its weekly budget, however, was 5,266 million tokens, against 6,579 million, a drop of 20%. SemiAnalysis puts the monthly API-equivalent value at $9,385 for that account, against $11,726 for the other.

SemiAnalysis chart comparing two Claude Max 20x accounts, with a 19% larger five-hour window but a 20% smaller weekly budget on one

Image: SemiAnalysis

The weekly figure is the one that matters most. The five-hour window resets many times in a week, so SemiAnalysis treats the weekly budget as the cap on a month. Heavy users who hit the weekly limit would lose out. Someone who only runs into the five-hour limit could see the same change as a gain.

SemiAnalysis measures limits by running repeated calls and watching how far each provider's usage meter moves. Its October 5 report sets out the method, though much of the data sits behind a paywall. That report also described catching an A/B test at an unnamed provider, where one of three identical accounts had about 20% lower limits. The provider called it an "extremely tiny" test of how to balance when people hit limits. Because the provider was not named, it cannot be tied to Anthropic.

Anthropic has not confirmed a test, based on what we could find. Earlier on October 22, Anthropic's Boris Cherny told a user "No changes on our end." Lydia Hallie, who works on Claude Code, said "our limits haven't changed." Both posts came before SemiAnalysis published, so they do not answer its specific claim. Anthropic's Max plan page says it may limit usage in other ways, including weekly and monthly caps, at its discretion.

Treat the findings as unconfirmed. Two accounts is a small sample, Anthropic has not verified a test, and many users saying their limits feel lower does not prove a cause.

How to check your own limits

Open Settings and then Usage in Claude to see your five-hour and weekly meters. Note when you hit each one over a week and keep dated screenshots. If the weekly limit is the one that stops you, that is the number to watch.

Comments

Reader support

No ads. Just independent reporting.

Techcity Authority runs without advertising. Hosting, tools, and the hours behind each piece are supported by readers instead. Even a small contribution goes a long way — and every supporter is credited by name below for the month they give.

Support the site

Monthly via PayPal · cancel anytime

Secure checkout on PayPal. Your name appears here for the month you support.

This month’s supporters

    No supporters yet this month — yours could be the first name here.