Immediate mitigation
- Stop automatic retry storms.
- Add exponential backoff with jitter.
- Reduce concurrency and batch non-urgent work.
- Shorten prompts and requested outputs.
Gemini API
Fix Gemini API 429 quota exceeded.
A 429 usually means your project is sending too many requests, using too many tokens, hitting a daily quota, missing billing, or retrying too aggressively.
API checklist
Reduce load, then verify quota.
Project checks
- Open Google Cloud quota pages for the project.
- Check daily quota, RPM, TPM and billing status.
- Confirm the model and endpoint match your quota.
- Check whether free-tier or regional limits apply.
Logging
- Log status code, model, endpoint and latency.
- Record retry count and backoff delay.
- Store response body without exposing API keys.
- Separate user-facing traffic from batch jobs.
Related