Same $200 Claude Plan, Two Limits: One Gets 20% Less a Week

Anthropic appears to be A/B testing the usage limits on its Claude Max 20x plan, the research firm SemiAnalysis said on X on 22 October 2026. On two Max 20x accounts measured the same day with the same model, one got a 19% larger five-hour window but a 20% smaller weekly budget, “so a heavy user gets 20% less from it over a month,” according to the chart SemiAnalysis posted.

“This only affects SOME accounts,” SemiAnalysis wrote. These are its measurements, not a change Anthropic has announced. We did not find a public response from Anthropic as of 06:35 UTC on 23 October on its newsroom, its help pages on usage limits and on the Max plan, or its status page.

Business Pill · RATE LIMITS

A short Business Engineer explainer on rate limits: why an AI product caps how much each user can do in a period. It teaches the general idea only and says nothing about any company in this piece.

What SemiAnalysis Measured

• The claim. “We detected an A/B test going on that DECREASES your monthly API-value by ~20% while INCREASING your 5h usage limits by ~20%,” SemiAnalysis wrote, adding that users on X reporting lower limits “ARE NOT WRONG.”
• The accounts. The chart, titled “Anthropic appears to be A/B testing Claude Max 20x limits,” compares two Max 20x accounts measured with Opus 5.5 on the same day.
• The money. SemiAnalysis puts the monthly API-equivalent agentic value at $11,726 for Account A, 59x the $200 price, and $9,385 for Account B, 47x.

Figures from the chart SemiAnalysis posted on X on 22 October 2026 (source line: SemiAnalysis Tokenomics Model). Both accounts are on Claude Max 20x, measured the same day with the same model. The dollar gap is our arithmetic.

The post is the first of a two-part thread; we read the first post and its chart.

What Was Posted on X

SemiAnalysis posted this on 22 October 2026 at 21:36 UTC. It says it detected an A/B test that lowers the monthly API-equivalent value of some Claude accounts by about 20% while raising their five-hour usage limits by about 20%, and that users reporting lower limits are not wrong. Its attached chart compares two Claude Max 20x accounts; the piece’s table takes its figures from that chart.

Posting on 23 October, @kimmonismus relayed SemiAnalysis’s finding that Anthropic is A/B testing Claude Max limits, roughly 20% more usage per five-hour window but 20% less per week for some accounts, and called it a step back for Anthropic. The poster’s reaction is theirs; the figures are SemiAnalysis’s.

Two Limits, in Plain Words

Anthropic’s own help page says paid plans have two usage limits: “a session limit that resets every five hours, and weekly limits that reset once every week.” Its Max plan page says Max 20x “includes 20 times the Pro plan’s per-session usage allowance” and that Max plans “also have a weekly usage limit that applies across all models.”

SemiAnalysis explains why the weekly figure is the one that sets a month: “The 5-hour limit resets many times within a week, so a month is capped by the weekly limits.”

As we read it, the trade cuts two ways. A heavy user who runs into the weekly limit gets less over a month on the B-type account. Someone who works in short bursts and never reaches the weekly limit would only notice the larger five-hour window.

How SemiAnalysis Measures a Plan

In an article published on 5 October 2026, SemiAnalysis said it reads every usage meter a plan shows after every request and works out how many tokens of each type fit in each limit. It then converts that into an API-equivalent dollar value for one assumed agentic workload, based on its own usage ratios in September. A different mix of tokens would give different dollar figures.

The same article described an earlier case at an unnamed provider, where one of three identical subscriptions had about 20% lower limits. SemiAnalysis wrote that the provider told it the account was part of an “extremely tiny” A/B test, that it “didn’t just decrease limits wholesale” and that it was testing “how to better balance when people hit limits.” The article does not name the provider. SemiAnalysis drew this conclusion: “It proves providers can silently change subscription limits at any time.”

What Anthropic’s Terms Say

Anthropic’s Max plan page says: “Price and plans are subject to change at Anthropic’s discretion.” It adds that “to manage capacity and ensure fair access to all users, we may limit your usage in other ways, such as weekly and monthly caps or model and feature usage, at our discretion.” The pages we read do not describe how limits are tested.

Connecting the Coverage

Our 11 October piece on SemiAnalysis’s numbers reported its estimate that subscriptions make up about 10% of Anthropic’s overall revenue but can take up over 40% of its inference compute. SemiAnalysis calls those figures rough numbers, not company figures. What’s new: the firm now reports two different sets of limits on the same plan.

Our law-firms piece reported on 22 October that, according to Bloomberg Law, law-firm attendees said Anthropic would not negotiate key privacy and data-security terms without a minimum spending commitment. That piece added a short Update on this same SemiAnalysis post; this piece sets out the full figures.

Our prospectus piece reported that Anthropic generated around $16.3 billion in revenue in the first half of 2026 and spent nearly $9 billion on computing capacity and infrastructure, according to the draft prospectus as reported by The Wall Street Journal. Our Spark piece reported Bloomberg’s report that Anthropic is expected to publicly file IPO paperwork in the coming days.

The Structural Read

As we read it, the weekly budget is the number that matters for heavy users: SemiAnalysis says a month is capped by the weekly limits, because the five-hour limit resets many times within a week.

As we read it, the dollar figures are a way to compare plans, not money paid or refunded: SemiAnalysis converts limits into an API-equivalent value for one assumed agentic workload.

As we read it, Anthropic’s own terms leave room for this: its Max plan page says price and plans are subject to change at its discretion and that it may limit usage in other ways, such as weekly and monthly caps.

Three Implications

FOR HEAVY USERS SemiAnalysis says the B-type account gives a heavy user 20% less over a month, because the weekly budget is 20% smaller.

FOR BURSTY USERS As we read it, someone who never reaches the weekly limit would only feel the 19% larger five-hour window.

FOR ANTHROPIC SemiAnalysis says only some accounts are affected; Anthropic has not announced a change, and we did not find a public response from it as of 06:35 UTC on 23 October.

The Business Engineer Lens

This story maps onto the Business Engineer framework The Subsidized AGI Economy.

The framework’s starting point: “The subscription is not one product. It is three assets bundled at a single price.”

As we read it, a limits test is how a lab adjusts what that bundle delivers without changing the $200 price: SemiAnalysis’s two accounts pay the same and get different weekly budgets. The framework was published in June 2026 and does not discuss this test.

What Is Not Established

Anthropic has not confirmed a test; the figures are SemiAnalysis’s measurements of two accounts. The sources we read do not say how many accounts are in the reported test; SemiAnalysis says only some are. The dollar values depend on the workload SemiAnalysis assumes. We read the first post of SemiAnalysis’s thread and its chart, its 5 October article and Anthropic’s help pages; we did not contact Anthropic or SemiAnalysis. Nothing here is investment advice.

The Bottom Line

SemiAnalysis says two Claude Max 20x accounts measured on the same day had different limits: one got a 19% larger five-hour window and a 20% smaller weekly budget, which it values at $9,385 a month against $11,726. It says only some accounts are affected; Anthropic’s terms say plans and limits can change at its discretion.