Anthropic A/B Tests Claude Subscription Limits on Some Accounts

Anthropic is running an A/B test on Claude subscription usage limits affecting only some accounts. Test-group accounts receive about 20% higher 5-hour limits while their monthly API-equivalent value drops by roughly 20%, according to SemiAnalysis.
Event version 1.
Key insights
Partial account rollout
The A/B test affects only some Claude subscription accounts, not all users. Accounts in the test group see different limit settings than other accounts on the same plan.
Higher 5-hour limits
Test-group accounts receive about 20% higher usage limits over the 5-hour window. This raises short-session capacity for affected users.
Lower monthly API value
The same test-group accounts see their monthly total usage, measured as API-equivalent value, decrease by about 20%. The change shifts capacity toward short windows and away from monthly totals.
User reports confirmed
Many users reported lower usage limits, and the reports correspond to the detected A/B test. The affected accounts are a subset of subscribers.