Claude Max users sue over a 20x claim quoted against the five-hour window

The decision at the centre of this is small and specific. Someone already paying $100 a month opens the pricing page, reads 5x and 20x, doubles the spend to $200, and waits for the ceiling to move ^[2]. The multiple describes usage inside a session window that resets every five hours ^[4]. The thing that stops them on a Thursday afternoon is the weekly limit, added in August, months after Max shipped in April 2025 ^[4][1].

Users actually read a multiplier as applying to the period they are billed for. One commenter quoted by Engadget said they thought they were buying 20x weekly usage "because weekly usage is what I'm paying for, not a multiplier to a session limiter within the service," and only learned otherwise from a Reddit post ^[5]. Teams tell themselves the caps are documented. They are, on a Claude support page that Engadget dates to a most recent update of August 7, 2026 ^[6].

The arithmetic the plaintiffs are working from is worth doing slowly, because it is the plan comparison every buyer runs. Take the advertised tiers at face value and the step from 5x to 20x is 4x more usage ^[11], for a price step of 2x ^[12]. The viral Reddit screenshot cited by Engadget claims that once the weekly cap is counted, the real gap is 1.7x ^[8]. Treat that as a user estimate rather than a measurement, and it still moves the answer: 2 divided by 1.7 means the $200 tier costs roughly 18 percent more per unit of weekly usage than the $100 tier ^[13]. Against Pro at $17 a month billed annually, the $200 plan is about 11.8 times the price ^[14].

Engadget's account describes a disclosure claim about how the 5x and 20x figures were presented, and does not say what Max's terms provided about introducing new usage limits later ^[15]. Anthropic had not responded to Engadget's request for comment at the time of publication ^[10]. The complaint puts the placement of the caps on trial, not Anthropic's right to retune them.

For anyone else selling metered access, that narrows into a usable check. For each paid tier, write down which limit binds first for the user you are actually selling to, then ask whether the advertised ratio holds on that limit. That check sorts into one of four results. The ratio is quoted on the binding limit and holds, the one clean outcome. It is quoted on the binding limit and does not hold, which is a wrong number and a cheap fix. It is quoted on a non-binding limit that happens to track the binding one, which survives until you change either meter. Or it is quoted on a non-binding limit that does not track the binding one, which is the case the Claude plaintiffs describe.

The operational rule that falls out of it: when you add a second meter after launch, re-derive the tier ratio on the new meter and print it where the buyer sees the price. A support article the buyer reaches through several hyperlinks is where Monica Vaca, one of two attorneys for the plaintiffs, is aiming her argument ^[7].