What if I fire a lot of AI requests at the same time?

Concurrent paid requests are capped per account; extras get a brief 429 with a Retry-After and can be retried once a slot frees.

We cap how many paid requests run at once per account to keep spending predictable. Extra concurrent requests get a brief 429 with a Retry-After header — retry after it, possibly more than once while the account is still at its cap; free-model requests aren’t limited this way.