Gemini API 429 RESOURCE_EXHAUSTED ('You exceeded your current quota' / 'Resource has been exhausted'): rate limit or billing, and how do I fix it?

Cause: 429 RESOURCE_EXHAUSTED means a quota for your Google Cloud project was hit: requests per minute, tokens per minute, or requests per day for that model. Limits are per project (not per API key) and per model, and the free tier's limits are much lower than paid tiers. Preview and experimental models often have tighter limits than stable ones. How to check: - Read the error details: they usually name the exceeded quota metric (per minute or per day, input tokens or requests) and the model. - In Google AI Studio, look at the project's usage and rate limits, and confirm which tier it's on. In Vertex AI, check Quotas in the Cloud console for the model and region. - A per-day limit resets daily; a per-minute limit clears within a minute. Fix: 1. Retry with exponential backoff and jitter. The official Google Gen AI SDKs retry transient errors (429, 503) automatically; don't stack your own tight retry loop on top. 2. Reduce pressure: lower concurrency, batch work (the Batch API has separate, larger limits), cache repeated prompts with context caching, and trim prompt tokens. 3. Raise the limits: link a Cloud Billing account to move off the free tier, or request a higher tier or quota increase. Spreading load over many free projects to dodge limits breaks the terms; don't. 4. Don't retry 400 or 403 errors: those are bad requests, invalid keys, or billing problems (for example a free-tier-restricted project), which need a manual fix. After: watch the usage dashboard, and put a request queue with a rate limiter in front of the API in multi-agent setups so all agents share one budget. Sources: https://ai.google.dev/gemini-api/docs/troubleshooting https://ai.google.dev/gemini-api/docs/rate-limits https://cloud.google.com/vertex-ai/generative-ai/docs/error-code-429