Gemini 3.8 Live is the default option for most low-latency voice agent experiences and real-time dialogue without reasoning-induced delays. It supports interleaved reasoning, asynchronous function calling, full session client content updates, and built-in audio streaming. Try in Google AI Studio
Documentation
Visit the Live API guide for full coverage of features and capabilities.
gemini-3.8-live
| Property | Description |
|---|---|
| Model code | gemini-3.8-live |
| Supported data types | Inputs Text, images, audio, video Output Text and audio |
| Token limits[1](https://ai.google.dev/gemini-api/docs/tokens)^ | Input token limit 131,072 Output token limit 65,536 |
| Capabilities | Audio generation Supported Caching Not supported Code execution Not supported File search Not supported Function calling Supported Grounding with Google Maps Not supported Image generation Not supported Live API Supported Search grounding Supported Structured outputs Not supported Thinking Supported (interleaved reasoning) URL context Not supported |
| Consumption options | Batch API Not supported |
| Versions | Read the model version patterns for more details. - Stable: gemini-3.8-live |
| Latest update | September 2026 |
| Model card | Model card |
Migrating from Gemini 3.1 Flash Live
Gemini 3.8 Live delivers ultra-low latency audio-to-audio interactions
and expands support for asynchronous workflows. When migrating from
gemini-3.1-flash-live-preview, review the following updates:
- Model string : Update your model string from
gemini-3.1-flash-live-previewtogemini-3.8-live. - Thinking level :
thinking_levelis not supported forgemini-3.8-live. Omitthinking_level(orthinking_config) from your session setup. - Asynchronous function calling : Async execution (
behavior: NON_BLOCKING) is now the default function calling mode. You can still use synchronous blocking mode for backwards compatibility by settingbehavior: BLOCKINGon your tool declarations. Function scheduling (SILENT,WHEN_IDLE,INTERRUPTED) is supported. - Client content updates :
send_client_contentis supported throughout the entire session lifecycle with explicit roles (userormodel). Settingturn_complete=trueunconditionally interrupts active model generation. If you send content withoutturn_complete, the server waits for subsequent messages before responding. - Proactive audio : Proactive audio is now permanently enabled. Setting
proactive_audio: falsereturns an error. - Affective dialogue : Affective dialogue is removed from the API. Remove any
enable_affective_dialogconfigurations from your code. - Turn coverage : Defaults to
TURN_INCLUDES_AUDIO_ACTIVITY_AND_ALL_VIDEO. Video frames are sent to the model by default, so only send frames when needed to manage context and cost. - Response modalities: Audio is the supported response modality. Enable output audio transcription if your application requires a text transcript.
For a feature comparison across all Live API models, see the Model comparison table.
[*] ↩︎