NewOpenAI- and Anthropic-compatible — no migrationRead the announcement
Reference

Rate limits

Limits vary by plan and by surface. The X-RateLimit-* and retry-after headers in every response tell you where you stand.

Rate limits & backoff

When you hit a limit, the gateway returns a 429 with a retry-after hint. Use exponential backoff with full jitter so a fleet of clients doesn't synchronize on the same recovery window.

Per-key limits are visible on the dashboard and can be raised on request for production workloads.