AI usage limits

Every page that answers “how many messages do I get?” with a single number is wrong within a month, because vendors change these constantly and tighten them under load. So these guides lead with the mechanic instead: what actually consumes a unit of quota, whether the window rolls or resets on a date, what happens the moment you hit the wall, and whether the cheaper fix is the next tier up or a different tool entirely.

ChatGPT limits

AI assistants

The cap is per-model rather than per-account, and free users are downgraded silently rather than blocked — which is why so many people report the product 'getting worse' mid-session instead of seeing an error.

Meters: Messages, per modelRolling hourly + monthly buckets

Claude limits

AI assistants

Token metering explains the whole confusing picture: two people sending the same number of messages get very different caps, because one pasted documents and the other typed sentences.

Meters: Tokens, not messagesRolling multi-hour, plus a longer-horizon cap

Cursor limits

AI coding tools

One instruction is not one request. The agent reads, plans, edits and verifies, and each step is billable — so a single refactor can spend a dozen requests before it reports back.

Meters: Agent requestsCalendar month, then usage-based overflow

Three things that are true of nearly every cap

Long conversations are the expensive part. Most of these products re-read the entire thread on every turn, so message forty costs many times what message two did. Starting a fresh chat when the subject changes saves more quota than any other single habit.
Rolling windows are not calendar days. A cap that restores gradually from your first message behaves nothing like one that clears at midnight, and mistaking the two is why people wait far longer than they need to.
The in-product number beats any published one. Vendors adjust allowances per-account under load. The banner telling you when your access returns is accurate; a table on a website, including ours, is a snapshot.

How these differ from our other guides

Pricing pages tell you what a plan costs. Limits pages tell you what you get for it before the product stops.
Comparisons help when you're deciding between two named tools.
Switch guides help once you've decided and need the migration mechanics.
Stacks help when one tool was never going to finish the job on its own.
Integration guides help when you already pay for two tools and want them talking to each other.

Every quota someone hits is a switching decision an assistant gets to answer.

People who run out of messages ask ChatGPT or Perplexity what to move to, and the answer comes from pages like these. If your tool has a more generous limit, that is worth nothing unless the assistants can read it. Check in sixty seconds.

Run a free AI visibility scan

📬 Get the best new AI tools delivered weekly

One concise email with fresh launches, trending picks, and featured standouts.