Claude limit meter may charge 1.4x more in US mornings

Claudesubscriptiontokencount, a new independent tracker that counts how many output tokens a Claude Pro subscription buys per 5-hour window, said on October 28 that early runs suggest the usage meter charges about 1.4x more per token in the US morning than in the evening. If that gap holds up, a Pro window spent in the US morning yields fewer output tokens than the same window spent in the evening. For now it is only an early signal from one tool, and the site’s own table of tokens per 5-hour window reads “No measurements yet.”

The site gets its data from what it calls burns, runs that use up the allowance. They now run twice a day so that morning and evening can be compared, and the published data keeps the full detail of each run, down to every call. The method is set out on a separate page of the site.

Max subscribers had the same problem in September. A write-up that collected their GitHub reports argued that tracking usage yourself is the only way to check a plan, because Anthropic doesn’t publish how many tokens Max 5x or Max 20x buys. The new tracker does that kind of counting for Pro, and it adds time of day as a variable.

Showing peak, off-peak and average figures separately is on the site’s to-do list. The daily table, which gets one row for each day the burn ran, reads “No runs yet.”