Gemini Partial Outage

Duration:

Jan 21, 2026, 9:00 PM UTC – Jan 22, 2026, 12:44 AM UTC

Description of the Issue:

Multiple features across the ElevenLabs platform that relied on Gemini models experienced elevated error rates: agents post-call conversation analysis, voice design, video generations, and others.

Root Cause:

Google Cloud Platform had an incident that caused throttling of Gemini models. According to them, this led users to switch to other models available through Google Cloud (Gemini 2.0, Claude, etc.), which in turn caused further throttling of those models due to unprecedented demand.

Timeline & Actions Taken:

• 21:00 UTC: Gemini services increasingly start returning 429 errors according to our logs.
  

• 21:43 UTC: The issue is flagged by customer support, and an incident is started. The engineering team begins investigating.
  

• 22:00 UTC: We start working on a fix to route traffic to a different LLM provider. Because no incident is posted on the GCP status page, we file an emergency ticket with their support.
  

• 22:28 UTC: GCP confirms there are throttling issues with all Gemini models, and their SRE team is working to resolve them.
  

• 22:54 UTC: Error rates drop significantly before the fix is released to production.
  

• 00:44 UTC: After monitoring the error rate we decide to mark the incident as closed.

Preventative Actions & Learnings:

Many, but not all, parts of our platform implement automatic fallbacks for LLM providers. We will prioritize bringing fallback coverage to 100% of our features.