[Groq docs]: when you exceed rate limits the API returns 429 Too Many Requests. Only 429 responses include the retry-after header (seconds to wait); every response carries x-ratelimit headers showing remaining requests/tokens and reset times for both the request and token windows. Limits apply at the organization level and free tier is much lower than paid. Check your actual quotas at console.groq.com/settings/limits. Back off on retry-after, and queue or throttle client-side to stay under both RPM and TPM.

Context: Groq enforces requests-per-minute and tokens-per-minute simultaneously. Hitting either returns 429.