Rate limit

Providers enforce limits along several axes at once: requests per minute (RPM), tokens per minute (TPM), and sometimes tokens or requests per day. Exceeding any of them returns an HTTP 429 error, and production clients are expected to retry with exponential backoff.

Limits typically scale with usage tier: new accounts start small and gain headroom as cumulative spend or verification level increases, with custom limits negotiated for enterprise contracts. For high-volume applications, available TPM can matter as much as price, since a cheap model you cannot call fast enough may be worthless.

Common strategies for living within limits include request queuing, spreading load across providers or regions, and moving non-urgent work to batch endpoints, which are governed by separate and much larger quotas.