Limits

Rate limits by plan, request size, context length and beta availability on the Ertas AI Gateway, and how to plan for them in your app.

Ask AI about this page

On this page

Rate limits

Each plan has its own API rate limit, and Max has the highest. An API call over the limit gets 429 rate_limit_error; back off and retry. If you need more, talk to us.

Request size and context length

Each model's context length is on its page in the app. A request larger than the model can accept gets 413 request_too_large.

Keys

Up to 25 active API keys per account. See API keys.

Availability during the beta

API inference is in beta, and we don't recommend it yet for intensive or stable production use. During the beta:

A model can refuse an API call when it's at capacity, with 503 and a Retry-After header.
A model can restart, which can take a few minutes. API calls in flight fail and new ones get 503 until it's back.

Design for both with retries (see Errors and retries). If you're planning production volumes, talk to us about the models, volume and latency you need.