Since yesterday I have an unexpected 429 resource exhausted error on my account for the E5 API Service for intfloat/multilingual-e5-large-instruct-maas:
I am nowhere near quota limit, so I was wondering if anyone knows what is happening
We’re seeing exactly the same issue on project (PII Removed by Staff). 100% of requests to
intfloat/multilingual-e5-large-instruct-maas now return 429
RESOURCE_EXHAUSTED, including single-request calls at <1 RPS — so it
is not caused by bursty usage. Quota usage shows nowhere near the
limit in the Cloud Console.
From our Cloud Logging data:
2026-04-13: 188 × 200, 0 × 429 (healthy)
2026-04-14: 25 × 200, 438 × 429 (breaks mid-day)
2026-04-15 – 2026-04-18: 0 × 200, 2200+ × 429
The global endpoint returns 500 INTERNAL for the same request.
Other Vertex AI APIs (Gemini embeddings, generateContent) work
normally from the same project and service account.
I’ve filed an issue tracker bug at in case that helps get
visibility. Has anyone from Google been able to look at this?