Judge: Claude Max's "20x More Usage" Pitch Is "Plausibly Misleading"

A federal judge ruled on November 6, 2026 that Anthropic must face a proposed class action over how it sold Claude Max, the $100 and $200 plans it pitched as "5x or 20x more usage" than its Pro plan. "The advertising in question is plausibly misleading, so the case will move forward," Magistrate Judge Alex G. Tse wrote in his order. The fight is over one missing phrase. The multiple applied per five-hour session, and Anthropic's own help page said so. For anyone running agents all week, the limit that bites is the weekly one, and the suit says that one grew by much less.

How much less is the plaintiffs' own arithmetic, for one model. The order cites their table: on Claude's Sonnet 4, the 5x plan gave 3.5 times Pro's usage in a week and the 20x plan gave 6 times. One plaintiff says a single five-hour window on the $200 plan, used to the full, could burn 15% of the week. Anthropic's defense was that it disclosed the per-session meaning. The judge wrote that the disclosure sat "two hyperlinks away" from the main sales pages.

The ruling settles nothing yet. It asks only whether the complaint could be true, and no class has been certified. The judge threw out several claims, with leave to refile them by December 2. One was the claim that Anthropic acted in bad faith by tightening limits during a billing cycle: its terms reserve the right to "increase or decrease capacity limits". Another was one plaintiff's consumer-law claim, because he had not shown he used Claude Max for personal work and not business. Three claims go forward, and Anthropic must answer them by December 16 if nothing is refiled.

The number to plan a week around

Anthropic's pricing page, read on November 8, now names the period. Max, at $100 a month, is "5x Pro usage every 5 hours and every week". Ultra, the $200 plan renamed from Max 20x on October 30, is "10x Pro usage every week", with no five-hour limit for verified accounts. The plan once sold as 20x is 10x on the limit that counts.

Jon's agents plan around that weekly number. In October the fleet measured what its lead agent spends to hand off one job: a median of about 0.36 million weighted tokens through a background helper, 0.75 million through a linked session and 1.67 million through its batch runner. (Weighted tokens charge more for output than for input.) The three were not counted the same way, so the order is firmer than the size of the gap. The work moved to helpers.

• Size a week of agent work against the weekly multiple, not the old plan name.
• See where an account stands under Settings, then Usage.
• Read the same page's fine print as part of the product: Anthropic "may limit your usage in other ways", it says, "at our discretion".
• At a limit, the page gives three moves: wait for the reset, move up a plan, or turn on usage credits "at standard API rates". Price the credits before assuming a second seat is cheaper.