Project: Unloadin (Tier 3) Model: gemini-3.1-flash-live-preview (Rate Limit page lists it as Gemini 3 Flash Live, Live API) API: Gemini Developer API, Live WebSocket (bidiGenerateContent) Client: LiveKit Agents Google Realtime plugin, one Live session per phone call
We place about 3 to 5 calls per minute, with 3 concurrent calls in production. Each call opens one Live session. Audio stays on that socket. It is not a separate HTTP request per audio frame.
During a call, about 25 seconds in, Gemini closes the socket:
google.genai.errors.APIError: 1011 None. Resource has been exhausted (e.g. check quota).
websockets.exceptions.ConnectionClosedError: received 1011 (internal error) Resource has been exhausted (e.g. check quota).
The greeting has already played. After the 1011, the call stays connected and the rest of the call is silent. The LiveKit plugin then retries in 0.1 seconds (max_retries 3), and those retries also get 1011. This hits a few calls in a batch of 30 to 40, not every call.
Rate Limit for Gemini 3 Flash Live on this project:
- Peak RPM: 7 / Unlimited
- Peak input TPM: 50.61K / 1,000,000
- Peak RPD: 206 / Unlimited
The TPM line stays near the bottom all month and never approaches the 1M line. The Spend page shows Tier 3 , billing is not setup , currnelty on pre paid,
Is there a concurrent Live session cap for gemini-3.1-flash-live-preview that this dashboard does not show?