429 — one of five limits was hit

The request was rejected by a limit. The action was not performed, no order was created, and no money was charged. For three of the five sources the body is identical and useless: {"success": false, "message": "Too many attempts. Please wait a moment and try again."}.

FIVE independent mechanisms respond with 429, and each one is resolved differently: the key rate limit (request frequency), the key token quota (generation volume), the subscription usage limit, the limit on concurrently active tasks, and the cooldown after a series of content blocks. Blind exponential backoff helps only with the first two; on the concurrency limit and the policy cooldown it just burns time.

Triage by two signals, both visible in the response itself. First — whether the body has a reason: no reason means the key rate limit. Second — whether there is a Retry-After header: it comes ONLY with token_quota_exceeded and policy_cooldown_active. The combination of these two signals distinguishes all five sources unambiguously.

GET /usage — the subscription limits in effect and current consumption, including the entry with limit_type: "concurrent". The limits of the key itself are not included: they are in the developer section.

• Guide: limits and the 429 code
• token_quota_exceeded — generation volume
• limit_exceeded — subscription usage limit
• concurrency_limit_exceeded — task slots
• policy_cooldown_active — policy cooldown
• GET /usage
• Account limits
• My key limits