Anthropic promised 20x more usage. Then developers hit a weekly ceiling.

Anthropic’s top Claude Max plan comes with a big promise: 20 times the usage of its $20 Pro tier. But developers paying $200 a month can still run into a separate weekly ceiling, and an expanded class-action lawsuit filed Tuesday says that wasn’t made clear enough.

The dispute is really about what “20x” is supposed to mean. Anthropic says the Max 5x and 20x multipliers apply per session, with limits resetting every five hours. But that’s only part of the story. The company also sets a weekly usage cap across all models and says it can add other restrictions to manage capacity. For Claude Code users, that matters because interactive coding sessions count against the same plan limits.

According to the complaint, which The Verge first reported, Anthropic added weekly limits in late July 2025, months after Max launched in April, while continuing to market the plans around the 5x and 20x claims. The plaintiffs say those limits were not adequately disclosed at signup. Anthropic has argued in a motion to dismiss an earlier version of the case that customers could find the information through hyperlinks during purchase, comparing that to a product label on the back of a package.

The larger problem is bigger than one company. AI coding work is wildly uneven: a quick fix may barely dent a plan, while a harder job can burn through it fast. That makes simple monthly pricing slippery, because a usage multiplier tells you how much more you get than Pro, not how much real work you can finish before the meter stops.

OpenAI’s Codex runs into the same basic issue from another angle. Its docs say usage depends on the task, the model, and where the work runs, and that it can draw from a shared allowance with other agentic products. When that runs out, users can buy more credits. OpenAI has also tested outcome-based pricing with some enterprise customers, charging only when an agent completes a task instead of billing raw compute. Different approach, same message: AI subscriptions are getting harder to explain because the work itself is hard to predict.

If the plaintiffs win, other providers may have to spell out their limits more clearly at signup, especially when usage resets and extra restrictions kick in. And that may be the real test here: not whether a plan says “20x,” but whether anyone can tell what that buys before the invoice arrives.