SemiAnalysis says Anthropic is A/B testing Claude usage limits; users report lower limits, not confirmed by Anthropic

On October 22, the analyst firm SemiAnalysis posted on X that Anthropic is running an A/B test on usage limits. It says many users have reported lower limits that day and that "they are not wrong". According to SemiAnalysis, which says it detected the test itself, the variant it saw cuts the monthly API value of a plan by about 20% while raising the 5-hour usage limit by about 20%, and it affects only some accounts. The post includes a screenshot.

This is a claim from SemiAnalysis and from users on X. Anthropic has not confirmed it. The post had about 1,100 likes and 71 replies by October 22.