Allowlist request — Gemini 3.x Live (bidi) preview models on project selldo-voice-ai

Gemini API
,
1

Hi ,

We run a production real-time voice-AI calling platform on Vertex AI and would
like to request allowlist / product-gate access to the following Gemini Live
(bidirectional streaming) preview models for our Google Cloud project.

Project ID: selldo-voice-ai
Billing account:
Primary region: us-central1 (also requesting: global)
API surface: Vertex AI Live API — LlmBidiService/BidiGenerateContent (WebSocket)

Models requested (Vertex publisher IDs):

  1. gemini-3.8-live
  2. gemini-3.8-live-extended-thinking
  3. gemini-3.1-flash-live-preview

Observed behavior (why we believe this is a gate, not a config issue):

  • Unary catalog GET publishers/google/models/gemini-3.8-live returns launchStage=GA
    at us-central1.
  • Opening a Live session (BidiGenerateContent) for the same model returns:
    1008 “Publisher model projects/selldo-voice-ai/locations/us-central1/
    publishers/google/models/gemini-3.8-live was not found or [no access]”.
  • Same 1008 at location=global (model not published there).
  • The model does NOT appear in ListModels (publishers/google/models), while
    GA models we already use do.
  • Reproduced with google-genai 1.71.0 (vertexai=True) and 2.23.0
    (enterprise=True); service-account auth (ADC) is correct and other Gemini
    models on this project work.

Stack: Python google-genai + google-adk (Vertex AI backend), service-account
auth. GA Live model gemini-live-2.5-flash-native-audio already works for us at
us-central1, so the integration itself is proven; we only need access to the
3.x Live SKUs above.

Business context: <e.g., outbound/inbound AI calling, X concurrent calls,
region India/APAC; production traffic on Sell.do Voice AI>.

Could you add project selldo-voice-ai to the Live-API allowlist for these
models, or point us to the correct preview-access form? Happy to provide any
additional details.

Thanks,
Vinayak Katkar