Why does Claude run out of messages, and how many do you get?

Claude's limits work differently from a simple message counter, which is why hitting them can feel unpredictable. Anthropic does not publish a fixed messages-per-period number; instead each tier has a usage budget that is consumed based on how much work you ask for, meaning long messages, long conversations, and the heavier models burn through it faster than short exchanges on a light model.

That is why the same plan can give one person hundreds of quick exchanges and another far fewer: a user running Opus 4.8 or Fable 5 on long documents consumes budget much faster than someone asking Sonnet short questions. The limit resets on a rolling window rather than monthly, so hitting it usually means a short wait rather than being locked out for the month.

The tiers scale this budget: Pro provides roughly five times the free allowance, and the Max plans (5x and 20x) multiply it substantially for heavy users. If you consistently hit Pro's ceiling, especially running Claude Code or pushing long documents all day, that is the signal to move to Max rather than fight the cap.

To stretch any budget: default to Sonnet 4.6 for routine work and reserve Opus and Fable 5 for tasks that need them, keep context focused rather than dumping huge irrelevant files, and start fresh chats for new topics so old context does not inflate every message.