Repeated 503 responses from gemini-3.8-flash on free tier

Hello,

We’re testing a private, server-side paragraph organizer with billing disabled.

  • Model: gemini-3.8-flash
  • Endpoint: POST https://generativelanguage.googleapis.com/v1beta/models/gemini-3.8-flash:generateContent
  • Runtime: Cloudflare Worker, direct REST request.
  • Results: Two confirmed upstream HTTP 503 responses on October 26, 2026, including one controlled retry with unchanged configuration and small fictional input.
  • Timing: Exact attempt times were not retained. Both occurred before 3:44:52 p.m. EDT (19:44:52 UTC), our subsequent status-check time.
  • Provider request IDs: Not captured.

AI Studio displayed free-tier quota for this model. Google’s status page reported operational when checked, including the free-tier view. Generation requests are now paused.

Could you advise whether this suggests intermittent capacity issues and what minimal, non-sensitive diagnostics would help investigate? We would like to establish the cause before changing the model, credentials or billing.

Thank you,
Stephen

I had/have an issue with permanent 503’s on a copied project all day long. Same API: gemini 3.8 flash. Did not test with other model. Google reports 503 are not due to user’s actions. So… yeah. Nice weather we’re having.