Project ID: gen-lang-client-0949398102 (display name: gais-first)
Model: gemini-2.5-flash-lite
Region: us-central1
API: Vertex AI (vertexai=True), google-genai Python SDK
Every generate_content() call to gemini-2.5-flash-lite returns:
429 RESOURCE_EXHAUSTED
{‘error’: {‘code’: 429, ‘message’: ‘Resource exhausted. Please try
again later.’, ‘status’: ‘RESOURCE_EXHAUSTED’}}
This happens on every single attempt, including a minimal, isolated
test call (“Say hello in one word.”) with no other code involved,
and persists across multiple retries with long backoff (30s/60s/90s).
Diagnostics already completed, all passing:
- client.models.list() succeeds and lists gemini-2.5-flash-lite as
an available model for this project/region - IAM permissions confirmed: roles/owner on this project
- Billing confirmed: active postpay (Google Cloud Services) account
with substantial credit balance - Quota console checked: all Gemini/Vertex AI quota rows show 0%
current usage (i.e. not at any visible limit) - Application Default Credentials freshly re-authenticated via
gcloud auth application-default login - aiplatform.googleapis.com API confirmed enabled for this project
- Network/IPv6 connectivity issue ruled out (forced IPv4 via
socket.getaddrinfo, same result) - Switching only the model name to gemini-2.5-flash (same project,
region, credentials, code) succeeds immediately and consistently
Since every other variable is held constant and only the model name
changes the outcome, this looks like an access/allocation restriction
specific to gemini-2.5-flash-lite on this project — possibly related
to its announced October 16, 2026 retirement. Is there a known issue
or an allowlist step needed to restore generate_content() access to
this specific model?
My last attempt to access the model was at 15:44 IST.
Happy to provide additional logs/request IDs if useful.