The Claude Max “20x” lawsuit goes ahead. What’s confirmed about Claude’s limits

What the judge decided on November 6

On November 6, 2026, U.S. Magistrate Judge Alex G. Tse let the proposed class action over Claude Max’s limits go ahead. “Some of Plaintiffs’ claims have shortcomings as pled, but not all,” his order begins. “The advertising in question is plausibly misleading, so the case will move forward.”

The case, Kahn v. Anthropic PBC, in the Northern District of California, began with Karl Kahn’s complaint on June 14, 2026. It was later combined with a second suit, and in September Kahn and another subscriber, Solomon Smith, filed one complaint together. They say Anthropic sold Max 5x and Max 20x as offering “5x or 20x more usage than Pro” when the plans’ weekly limits weren’t 5 or 20 times Pro’s. By their own calculation for the Sonnet 4 model, Max 5x got 3.5 times Pro’s week and Max 20x got 6 times. Anthropic’s motion to dismiss argued that it “makes clear” the multiples apply “per session.”

At this stage a court takes the complaint’s account as true, and on that account Judge Tse found the per-session explanation too far from the sale to end the case: “Anthropic’s per session disclosure is located two hyperlinks away from the primary sales pages on claude.com.” The asterisk on the sales page, “Usage limits apply,” didn’t settle it either: “The disclaimer says nothing about usage limits per session.” Because Max is billed monthly, he wrote, buyers could reasonably read the multiples as 5 or 20 times Pro’s usage “per month.”

Claim	Ruling	The court’s reason
Smith’s claim under California’s Consumers Legal Remedies Act	Goes ahead	He’s a student, and “Educational use is plainly not commercial use”
Kahn’s claim under the same law	Dismissed, but can be amended	He didn’t show that he uses Claude for personal rather than business purposes
Negligent misrepresentation	Goes ahead	The same finding that the advertising is plausibly misleading
Breach of contract	Goes ahead	Anthropic’s terms leave a plan’s details to “the order process,” which plausibly included the “more usage” pages
California’s unfair competition and false advertising laws	Dismissed, but can be amended	The plaintiffs still subscribe, and didn’t explain why money damages wouldn’t be enough
Changing limits during billing cycles	Dismissed, but can be amended	The terms let Anthropic “increase or decrease capacity limits,” and “Anthropic didn’t tighten usage limits as a bait-and-switch”

The plaintiffs’ amended complaint is due December 2, 2026. If they don’t file one, Anthropic must answer the complaint by December 16.

This is a ruling on whether the claims can proceed, not on whether they’re true. It doesn’t find that Anthropic misled anyone, it doesn’t test the plaintiffs’ weekly figures, and it doesn’t certify a class. It also rules on the complaint as filed, so it doesn’t mention Anthropic’s October 30 rename or Mike Overell’s account, below, of how Max 20x’s weekly limit was set. Anthropic didn’t immediately respond to Courthouse News, and as of November 8 we haven’t found a public statement from it on the ruling.

What the ruling means if you pay for Max or Ultra

There’s nothing to claim yet. The plaintiffs propose a class of, in Anthropic’s summary of their complaint, “all natural persons in the United States who purchased or upgraded to a Max 5x or Max 20x plan from April 9, 2025 to the present.” The court hasn’t certified that class, and there’s no settlement or claim form. The original complaint asks for “actual damages, and/or restitution or disgorgement” for each member of the class, among other things. None has been awarded.

The ruling doesn’t change your plan. It orders nothing about limits. The claim closest to the mid-October reports below, that Anthropic changed Max during billing cycles, is the one the court dismissed, for now: it read Anthropic’s terms as reserving the right to change capacity, and found that in doing so “Anthropic exercised its contractual rights.”

The words on the sales page have already changed. Claude’s pricing page no longer says “5x or 20x more usage than Pro.” It now lists Max at “5x Pro usage” every five hours and every week, and Ultra at “10x Pro usage” every week, with “No 5-hour limit” for verified accounts. Anthropic hasn’t said whether the lawsuit had anything to do with the change.

Personal and business use are treated differently. The consumer-protection law whose claim survived covers buying “for personal, family, or household purposes.” Kahn’s claim under it failed for now because he didn’t say what he uses Claude Code for beyond “computer programming tasks,” and the judge noted he “could be an entrepreneur who uses Claude Code to develop software for sale.” Smith’s coursework counted as personal use.

What changed on October 30

On October 30, 2026, Anthropic renamed its Max plans: Max 5x is now Max, and Max 20x is now Ultra. “Max is still $100/month, and Ultra is still $200. If you’re a subscriber, there’s nothing you need to do.” In the same post: “We’re also removing the 5-hour limit for verified Ultra accounts, starting today with most existing subscribers,” and “Weekly limits remain unchanged.”

Mike Overell, who leads product for Claude subscriptions, explained the old name the same day. When the top plan launched in April 2025, “it had a single limit: 5 hours. It gave you 20x Pro’s limit, so we called it Max 20x. When we added weekly limits to Max 20x, we set them to 10x Pro’s (the 5-hour limit stayed at 20x).” He went on: “I get that some new users assumed 20x applied to both limits. We could have done better here, and I’m sorry.”

Here are the plans as Claude’s pricing page now lists them, with their old names:

Plan	Called until October 30	Price a month	Every five hours	Every week
Pro	Pro	$20, or $17 billed yearly	1x	1x
Max	Max 5x	$100	5x Pro	5x Pro
Ultra	Max 20x	$200	No limit once verified; was 20x Pro	10x Pro, as before

The multiples are relative to Pro, whose limits Anthropic still doesn’t publish, so they don’t tell you how much work a week holds. And the announcement’s “Weekly limits remain unchanged” doesn’t address the reports below.

Verification is Anthropic’s call. Its Ultra page says it verifies accounts automatically, “using factors like your account and payment history.” Until then, “a temporary limit applies,” and “Verifying your phone number at checkout is a separate step and doesn’t remove this limit.” Anthropic removes it automatically once the account is verified, “so there’s nothing you need to do.” Its pages don’t say how large the temporary limit is or how long verification takes.

Max and Ultra also have a separate weekly limit for Fable, Anthropic’s top model. Its help center says “You can use up to 50% of your weekly limit on Fable models,” and that past that point you can pay for more Fable with usage credits or switch to another model. On Pro, Fable runs on usage credits from the start.

What users have reported since mid-October

Since mid-October 2026, Claude subscribers have filed dozens of reports on Claude Code’s GitHub repository that their limits run out sooner than they used to, for the same work. They come from Pro, Max 5x, Max 20x and Team accounts, and several date the change to between October 15 and 17.

Some reporters measured it. They count the tokens in their own Claude Code transcripts, weight them by API price, and compare that with how far the weekly meter moved:

Report	Plan	What they measured
#103376, October 19	Max 20x	The weekly meter went from 0% to 54% in the 22 hours after the reset, on about the same tokens as the first 22 hours of the week before
A comment on #103376, October 19 and 26	Max	About 5.2 million weighted tokens per 1% of the week on October 5 to 9, and about 3.85 million from October 19: “a cut of about 30% at the same price”
A comment on #103376, October 25	Max 20x	About 1.1 to 1.25 billion weighted tokens a week until October 17, and about 640 million in the week after
#105240, October 26	Max 5x	94% of the week used in under 48 hours, all of it by Claude Code, figures Anthropic’s support confirmed for the account
#105352, October 27	Pro	About half the week used in two sessions over two days

The Max 5x reporter first noticed it “around October 17,” and says it got “significantly worse” in the week that began October 24. Others describe meters that rise while nothing runs: a Pro user on Linux (#105015) saw their five-hour meter reach 100% about three hours after it reset on October 25, with “zero API requests” in their Claude Code transcripts over that time.

The reports haven’t stopped, though they don’t all agree. On October 28, a Max 20x user wrote that about 60% of their week went in two days from October 27, and then: “After updating, consumption looks back to normal for the same kind of work.” On October 31, an Ultra subscriber wrote that their week had “lasted 6–7 days” for months, and since about October 24 “runs out after about 4 days, with the same models, same effort level and same workflow.”

Support hasn’t been able to settle it. The Max 5x reporter writes that Anthropic’s support confirmed the figures but said front-line support “does not have access to request/token history or tools that would allow them to compare the current week against previous weeks,” and sent them to GitHub. As of November 8, none of the reports we read has a reply from Anthropic’s staff on GitHub. Anthropic answered on X instead.

Not every report holds up

Measuring a limit from the outside is easy to get wrong. On October 24, a Max 20x user reported that a week held “about half the usage” of the week of October 8 (#104750). Within half an hour, they withdrew it: “the week of Oct 8 included a free weekly limit reset I had forgotten about, so it held about two weeks’ worth of usage.” Measured again, their recent weeks were “consistent with each other,” and they saw “no evidence of a reduced limit.”

Anthropic gives some plans occasional limit resets, which “set your usage limits back to full when you choose to use one,” so a week with one in it holds more than a week without. Other reports point to causes on the user’s side or inside Claude Code: subagents that lose their prompt cache while waiting on long commands (#103309), background agents that re-send their whole context each time a sleeping Mac wakes (#102597), and long sessions. Since November 4, Claude Code 2.1.318 moves a background subagent’s shell command to the background after three minutes, “so the prompt cache doesn’t expire while it runs.” Anthropic’s Claude Code documentation says a session open for hours “can use far more of your plan limits than your activity suggests,” because the conversation is sent again with every request, and the first message after a break “misses the cache and reprocesses your full context.”

The complaints aren’t new, either. Similar ones ran through September and early October, such as “Weekly usage limit consumption rate increased ~3.6x after September 25 reset” (#97398).

Headless Claude Code uses the five-hour window faster

One gap has been measured again and again. Since September 25, users in issue #97074 have reported that Claude Code run headless, with claude -p, which its transcripts record as sdk-cli, uses about 1.8 times as much of the five-hour window per token as the same work in the interactive terminal. A Team user measured 1.86 times on October 27. A Max 20x user later measured “about 2.5× as fast” as either the terminal or Claude Desktop’s Code tab, which is metered like the terminal, and found the weekly limit “probably not weighted” in a coarse, pooled look. On November 3, another Max 20x user, whose account no longer had a five-hour window, measured the week more finely for the Agent SDK’s label, sdk-ts, and found it used “1.14× to 1.32× faster” than other labels for the same work: “smaller than the 5-hour gap reported here, but it is not zero.” By November 8, the issue had 73 comments and still no reply from Anthropic.

The developer Theo posted on October 25: “All Agents SDK usage will drain your 5h limits faster than using Claude Code directly.” He wrote that it doesn’t affect weekly limits, and that he had reported it to Anthropic.

It matters if you use Claude Code through another app, because many apps run it this way. Paseo’s 0.14 beta changed the name its Claude Code sessions report to “Paseo,” and calls the change “a workaround, not an Anthropic-confirmed accounting fix.” A pull request for T3 Code, whose author measured “about 1.83x more work per point” with a new name, was closed unmerged: a maintainer cited the project’s rule that default changes need approval first, and noted that users can already set CLAUDE_CODE_ENTRYPOINT themselves.

Enjoy runs Claude Code headless too. So if these reports are right, Claude Code in Enjoy uses the five-hour window faster than Claude Code in a terminal, except on a verified Ultra account, which has no five-hour window. Its sessions carry the sdk-cli label, and the one pooled look at the weekly limit for that label found it “probably not weighted”; the finer November test measured the Agent SDK’s label, not this one.

What Anthropic has said

On October 22, Boris Cherny, who created Claude Code at Anthropic, replied to a user who said Claude Code usage was “draining 5x faster”: “No changes on our end.” Later that day he wrote to the same user: “We’re digging into this and will report back very soon.”

On October 28, he did: “Thanks to everyone who shared session IDs. We went through all of them. We found a few smaller things along the way that made some sessions burn through tokens faster than they should have, and we’re rolling out fixes to those right now.” He went on: “We didn’t find a single root cause or any regression in Claude Code. In most cases, users who saw faster usage against limits were traced back to one or two heavy sessions (xhigh effort, /goal, Workflows, Agent Teams, long SDK jobs).” His advice: “keep an eye on the effort level you’re on. Subagents inherit it by default and it can add up fast.”

He didn’t list the fixes or say which version carries them, and he didn’t mention the headless gap or SemiAnalysis’s test. Claude Code’s changelog has four changes since that bear on usage, none of them named as one of the fixes. Version 2.1.314 adds idle compaction to VS Code and Claude Desktop’s Code tab: “a 200K+ token conversation left idle is summarized, and the summary is used on your next message if the prompt cache has expired.” Version 2.1.316 adds CLAUDE_CODE_SUBAGENT_EFFORT, “to set the effort level subagents run at when nothing else names one, instead of inheriting the session’s.” Version 2.1.317 has subagents on a different model from the main conversation use “that model’s saved effort level or its default, instead of the session’s effort level.” And 2.1.318 makes the change to subagents’ long commands described above.

Earlier versions touched the meters, not the limits. Version 2.1.311, released October 26, fixed /usage “never showing large prompt-cache misses under ‘What’s contributing to your limits usage?’” and fixed Claude Code “retrying requests the API marked not to retry, which could hold a turn for minutes.” Version 2.1.312, on October 27, calls the five-hour window the “5-hour limit” instead of the “session limit.”

Anthropic’s S-1, filed October 26, says: “We have also changed, and may in the future change, usage limits, rate limits, or model and product capabilities.” It doesn’t say which changes it means.

Its help center now describes the limits with the new names. The Max page says Max “includes five times the Pro plan’s usage every five hours and every week,” and the Ultra page that Ultra “includes 10 times the Pro plan’s usage every week.” Weekly limits reset “at a fixed time each week that is assigned to your account.” Both pages, like Claude’s pricing page, keep the same reservation: “we may limit your usage in other ways, such as weekly and monthly caps or model and feature usage, at our discretion.” When Anthropic released Opus 5.5 on September 22, it said: “we’re increasing five-hour usage limits on Pro, Max, Team, and seat-based Enterprise plans.”

What SemiAnalysis reported

SemiAnalysis, a research firm that tracks the usage limits of AI plans, posted on October 22, 2026 that users reporting lower Claude limits “are not wrong.” In its words: “We detected an A/B test going on that decreases your monthly API-value by ~20% while increasing your 5h usage limits by ~20%. This only affects some accounts.”

Its chart compares two Claude Max 20x accounts, now Ultra, measured on the same day with the same model, Opus 5.5:

	Account A	Account B	B against A
Five-hour window	1,740 million tokens	2,076 million tokens	19% more
Weekly limit	6,579 million tokens	5,266 million tokens	20% less
A month at API prices	$11,726	$9,385	20% less

The token counts are for SemiAnalysis’s own mix of agentic work. The dollar figures price a month of that work at Anthropic’s API list prices, 59 and 47 times the plan’s $200 price. SemiAnalysis’s method reads the usage meters after every request and works out how many tokens fill each one. A month depends on the weekly limit, because “the 5-hour limit resets many times within a week, so a month is capped by the weekly limits.” It has published figures for Max 20x only, and its post doesn’t say how many accounts it measured.

There is a precedent in SemiAnalysis’s own reporting. Its October 5 article says “just 1 of 3 of the same subscription we tested for a particular provider had ~20% lower limits than the other 2.” The provider, which SemiAnalysis didn’t name, confirmed an “extremely tiny” A/B test on limits.

Paweł Huryn, who publishes his own measurements of AI plans, disagreed: “From what I see, nothing is being degraded or A/B tested.” He had published a 45.3 times multiplier for the $200 plan on October 1, about $9,060 a month at API prices and close to SemiAnalysis’s Account B, and wrote that the “decreased” value “is just what I originally measured.” His benchmark isn’t SemiAnalysis’s, so the two sets of numbers aren’t strictly comparable.

As of November 8, Anthropic hasn’t commented on SemiAnalysis’s report.

Who it would affect

If the users’ measurements are right, anyone who uses most of their week runs out sooner, on Pro, Max, Ultra and Team alike. Someone who rarely gets near the weekly limit wouldn’t notice.

SemiAnalysis’s test, if it’s real, traded a bigger five-hour window for a smaller week, so in its words “a heavy user gets 20% less from it over a month.” A verified Ultra account has no five-hour window left to gain, so on Ultra the week is what counts.

The October 30 change helps people who hit the five-hour limit on the $200 plan, once Anthropic verifies them. It leaves Pro’s and Max’s limits as they were, and doesn’t help anyone who runs out of the week.

The headless gap is separate. By the reports in #97074, it costs anyone who runs Claude Code through claude -p, the Agent SDK or an app built on them more of the five-hour window. One November test found the Agent SDK’s label using the week somewhat faster too, by 1.14 to 1.32 times. That applies to Pro, Max, Team and unverified Ultra accounts, which still have the window.

How to check your own limits

1. On claude.ai, open Settings > Usage. Anthropic’s help center says it shows when your weekly limit resets next. On Ultra, “If you don’t see a Current session bar, your account is verified.”
2. In Claude Code, run /usage. On a Pro, Max, Ultra, Team or Enterprise plan, it shows your usage bars and a breakdown of what used them, such as subagents and MCP servers, with suggested changes. Update to 2.1.311 or later, so it lists large prompt-cache misses too.
3. Watch the effort level. Cherny’s advice is to keep an eye on it, because subagents inherit it. From Claude Code 2.1.316, CLAUDE_CODE_SUBAGENT_EFFORT sets subagents’ effort separately, and from 2.1.317, a subagent on a different model uses that model’s own effort level.
4. Keep your own baseline. Claude’s meters show how much of each limit you’ve used, not how big the limit is, so they can’t tell you whether your limit changed. Note how much of the weekly meter a typical day of your work uses, and when you use a limit reset, and compare from week to week.
5. If you think your limits are wrong, tell Anthropic. Cherny asks for your /feedback in Claude Code “If something still looks off.” For support, sign in at claude.ai, select your initials or name in the lower left, and choose Get help, as Anthropic’s support article describes. One reporter was told front-line support can confirm your figures but can’t compare weeks, so bring your own numbers.

If the week runs out early

• Wait for the reset. Claude Code 2.1.234 and later can wait for the limit to reset, then continue the interrupted task.
• Use a limit reset. If Settings > Usage offers one, it sets your five-hour or weekly limit back to full, once. Unused resets expire.
• Pay for more. On Pro, Max and Ultra, usage credits in Settings > Usage let you keep working past the limit, up to a monthly spend limit if you set one. Anthropic says credits bought on or after December 7, 2026 expire 12 months after purchase, and those bought before then don’t expire.
• Move up a plan. Ultra has twice Max’s week, 10 times Pro’s against five, for $100 more a month, and Anthropic says “If you upgrade, your new plan’s limits apply right away.” Our plan guide compares Pro, Max and Ultra.
• Spend less of it. Anthropic’s documentation says Sonnet “handles most coding tasks well,” and suggests keeping Opus “for complex architectural decisions or multi-step reasoning,” lowering the effort for simpler tasks, and running /clear between unrelated ones.
• Use another account or agent. Enjoy shows the limits of every Claude account you’ve signed in to, and when each resets. When one runs out in the middle of a task, Enjoy moves the conversation to another account that still has usage, or waits for the reset and carries on, while your computer is awake with Enjoy open. You can also hand the conversation to Codex or another agent, on that agent’s own plan.

Enjoy doesn’t change Anthropic’s limits: those belong to your Claude plan. Its own price is on the Pricing page.

Anthropic has now said how its plans’ weeks compare, but not how big they are, and it hasn’t explained the mid-October reports, the headless gap or SemiAnalysis’s test. We’ll update this memo when it does, and as the lawsuit moves: the plaintiffs’ amended complaint is due December 2.