Max20x buys only 1.5x of Max5x, says one team’s tally

The developer behind Tokenism says a Claude Max 20x account now gets only about 1.5 times the weekly Claude Code capacity of a Max 5x account. When his team started tracking on August 23, the gap was about 2.5x. He published the tally on Reddit on September 22.
Both plans moved, in opposite directions. Measured in his normalized units, Max 5x went from 47 a week to 65, and Max 20x fell from 118 to 95. The cheaper plan gained weekly capacity while the more expensive one lost it. In another part of the post he puts the earlier Max 20x advantage at roughly 2.2x. The August 23 unit counts give about 2.5x.
The numbers come from Tokenism, a tool he built so his team could pool their Claude Code subscriptions. It sends the Claude review on GitHub to whichever account has the most usable capacity left. Before he built it, he had measured on his personal Max 20x account that one weekly limit was worth about six sessions, and then he saw posts on the subreddit reporting the same thing. That led him to measure it across the team.
The method scales each plan by its advertised session size. A Pro session counts as 1 unit, a Max 5x session as 5 and a Max 20x session as 20. He then tracks how much of the weekly allowance is consumed for a given amount of session usage, which converts each plan’s week into units. The weekly totals of the two Max plans, set side by side, give the multiple.
He calls the result observed Claude Code throughput. It rests on one assumption: that Anthropic gives each plan the session size it advertises. The bigger blind spot, in his view, is the unit itself. “What i cant measure is the fluctuation in the value of 1 unit, which is probably more impactful,” he wrote, and he has asked for suggestions on how to track it, floating a spare Pro account as one option.
Anthropic’s own published estimates never put the weekly gap between the Max tiers anywhere near fourfold. When it announced weekly caps on July 28, 2025, Anthropic estimated 140 to 280 hours of Sonnet 4 a week on the $100 Max plan and 240 to 480 hours on the $200 plan, with 15 to 35 and 24 to 40 hours of Opus 4. On Sonnet 4, both ends of the $200 range sit about 1.7 times above the $100 range. TechCrunch also reported that Anthropic described the $200 plan as offering 20x the usage of Pro, with limits triggered by token counts rather than by time.
Anthropic also changed the weekly limits during his tracking window. A temporary 50% boost to Claude Code weekly limits ran through September 13. On September 14 the standard limits were set 25% above the original baseline, which amounts to a 17% cut against the boosted level, for Pro, Max, Team and seat-based Enterprise plans alike. Anthropic’s follow-up post on that change said more usage changes are coming.
Up next
Xiaomi’s new MiMo models carry an unverified top-6 claim
Anyone pulling Xiaomi’s fresh MiMo-V2.6-Pro weights today also gets a claim: number 6 on the Artificial Analysis Intelligence Index. The ranking traces back to a post dated September 13, 2025.
Related stories
- Claude Code Ultra week measured at 17B Opus cache reads
- Claude limit meter may charge 1.4x more in US mornings
- Claude Max subscribers are auditing Anthropic’s meter
- Anthropic deleted the post that called a cut a 25% raise
- Chisle’s author says it cuts oversized tool output by 48%
- cc-limit-pacer drops Claude Code a model tier before lockout
Comments
No comments yet. Be the first.
Join the conversation
Sign in to leave a comment. It appears after moderation.

