Skip to main content

Plans and quota

Three plans. Anonymous needs nothing, Free needs a Google sign-in, Pro is a subscription. Current prices are on the pricing page; the limits below come from the code.

Limits

AnonymousFreePro
Rewrites3 per day40 per month300 per month
Words per request1,0002,5005,000
Input tokens per period9,000200,0001,250,000
Output tokens per request2,0484,0968,192
Rewrites at once112
Per minute3620
Model tiersstandardstandardstandard, fast, auto
Expand from notesyes
Custom templates12
Voice profiles1, measured only3, with the inferred half
Detect runs per day320100
Help questions per day1560200

Two meters, not one

A rewrite spends a request and input tokens, and you can run out of either.

The token meter is charged the whole prompt, not just your draft. The instructions alone are around 1,400 tokens before you have written a word. That is why the token allowance looks large next to the request count: on Free, 40 requests at roughly 1,400 tokens of overhead each is most of a 200,000 budget before any drafts are counted.

Practically: short drafts run out of requests first, long ones run out of tokens first.

Periods

Anonymous resets daily. Free and Pro reset monthly. resetAt on a 402 and on GET /api/usage tells you exactly when.

Credits

When the included allowance is gone, a signed-in account can keep going on credits bought in one-time lots.

  • Included quota is always spent first. Credits are the fallback, never the default.
  • The usage event on a rewrite stream says which paid, via fundingSource: included or wallet.
  • API keys spend the owning user's wallet. There is no per-key balance.
  • Credits are reserved when a rewrite starts and settled when it finishes, so a request in flight shows in reservedCredits.

Running out

What you hitStatusWhat clears it
Requests or tokens for the period402 quota_exceededThe period reset, or upgrading
Allowance gone and wallet short402 credits_requiredBuying credits
Too many at once429 concurrencyA rewrite finishing, seconds
Too many per minute429 rate_limitThe next minute
Draft too long413 word_capSplitting the draft

See Errors for the response bodies.

Model tiers

standard is the default and the only tier below Pro. Pro adds fast — quicker, cheaper, slightly less careful — and auto, which picks. There is no tier that changes the rules about facts.