Skip to main content
Your plan sets two limits for your account: the requests a minute, and the requests in flight at once. Every key on the account and your own use of the Morse app share both. Morse spaces the requests a minute evenly, so 60 a minute is one a second. An idle account can send its burst at once, which is 1 plus a tenth of its requests a minute. After that, Morse admits requests at the even pace. Every request counts toward the requests a minute, except GET /v1/models, which has its own limit below. The requests in flight count chat completions, image generations, web searches, and page fetches.

​
Fixed limits

These endpoints have their own limit for each account, the same on every plan:

​
Retry a 429

A request over a limit gets 429 and a Retry-After header, a whole number of seconds to wait.
  • rate_limit_exceeded means the account used its requests for the minute. Wait the seconds in Retry-After, then retry.
  • concurrency_limit_exceeded means the account has as many requests in flight as its plan allows. Retry-After is 1. Retry when one of your requests finishes.
Web search and page fetch also have a limit that all Morse accounts share. A request over that limit gets 503 with the code upstream_busy and a Retry-After header. Retry it the same way.