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

A Max 20x subscription used to give a team roughly two and a half times the weekly Claude Code capacity of a Max 5x. The developer who pools his team's subscriptions and has been logging the split since 23 August says the gap is now about 1.5x, and he posted the tally on Reddit.

At a glance

• The developer behind Tokenism, a tool that routes GitHub Claude reviews to whichever pooled account still has capacity, has been tallying weekly allowances across plans since 23 August.
• In his normalized units, Max 5x weekly capacity rose from 47 to 65 while Max 20x fell from 118 to 95, closing the gap from about 2.5x to 1.5x.
• The tally assumes Anthropic honours the advertised session limits, and it cannot see how much real work one unit buys, which he suspects matters more than the ratio itself.

If you have not been tracking the limit saga: according to Truefoundry, Anthropic cut Claude Code's 5-hour limits during weekday peak hours, 5–11 AM PT, from March 2026, and acknowledged on 31 March that users were hitting limits faster than expected. According to Explainx, the temporary 50% weekly boost, extended twice since July, expired on 14 September 2026 and was replaced the same day by a permanent 25% raise over each plan's baseline.

Max 5x gained 18 units, Max 20x lost 23

When he started on 23 August, a week on Max 5x came to 47 normalized units and a week on Max 20x to 118, about two and a half times more. The latest reading gives 65 and 95, a gap of roughly 1.5x. Both ends moved: the cheaper plan got better, the expensive one got worse.

He describes the earlier gap as roughly 2.2×, his own summary of the period rather than the 23 August reading, and the direction holds either way. The figures come out of live routing rather than a benchmark run, since Tokenism sends each GitHub Claude review to whichever pooled account has the most usable capacity left.

One Pro session counts as one unit

The normalization is his own, and it is simple. A Pro session counts as one unit, a Max 5x session as five, a Max 20x session as twenty, matching how the plans are sold. He then tracks how much of a weekly allowance one session's worth of usage eats, which turns two different budgets into one comparable number.

Think of it as measuring fuel tanks in tankfuls of the smallest car rather than in litres: fine for comparing tanks, useless if the fuel itself changes strength. Before building Tokenism he had measured on his own Max 20x account that one weekly limit worked out to about six sessions, and posts from other users lined up with that, which is what pushed him to measure across a team.

The weekly reset does not move, the 5-hour window does

Two clocks govern an account. The 5-hour window is rolling and starts when you send the first message of a session; the weekly limit resets at a fixed day and time assigned to the account and does not shift. Max plans carry two weekly caps, one across all models and a second scoped to heavier models such as Opus.

According to Truefoundry, that weekly cap counts active compute hours from a bucket shared with Claude.ai chat and Cowork; it rates Max 5x at $100 a month for roughly 140–240 Sonnet hours plus 15–35 Opus hours a week, and Max 20x at $200 a month for 240–480 Sonnet hours plus 24–40 Opus hours, as testing ranges rather than fixed counts.

According to SessionWatcher, how fast that budget drains depends on message length, attachment size, conversation length, tool use, model choice, effort level and artifact creation. Anthropic has said the limits exist because a small share of power users ran 24-hour sessions, shared credentials across teams and burned thousands of dollars of compute on $20 subscriptions, according to Truefoundry.

He names the hole himself: the tally assumes Anthropic honours the advertised session limits, and he cannot see how the value of one unit fluctuates, which he suspects matters more than the ratio. In our view the striking part is that both ends moved, with Max 5x gaining while Max 20x lost, so a plan sold at four times the session budget now measures at one and a half.

Watching the 125% baseline land

The next readings are the interesting ones. According to Explainx, the permanent 25% raise over baseline took effect on 14 September, which against the 150-point promo allowance works out to a reduction of about 17%. Whether that shows up as another slide in the 65 and 95 figures should be visible in a month of Tokenism logs. No date for a next update has been given.