Gemini 3.1 Flash Live Preview is a legacy preview model for real-time dialogue and voice-first AI applications.
[!NOTE] Note: We recommend updating to Gemini 3.8 Live as the default option for most low-latency voice agent experiences. See the migration guide.
Documentation
Visit the Live API guide for full coverage of features and capabilities.
gemini-3.1-flash-live-preview
| Property | Description |
|---|---|
| Model code | gemini-3.1-flash-live-preview |
| 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 URL context Not supported |
| Consumption options | Batch API Not supported |
| Versions | Read the model version patterns for more details. - Preview: gemini-3.1-flash-live-preview |
| Latest update | March 2026 |
| Model card | Model card |
Migrating from Gemini 2.5 Flash Live
Gemini 3.1 Flash Live Preview is optimized for low-latency, real-time dialogue.
When migrating from gemini-2.5-flash-native-audio-preview-12-2025, consider
the following:
- Model string : Update your model string from
gemini-2.5-flash-native-audio-preview-12-2025togemini-3.1-flash-live-preview. - Thinking configuration : Gemini 3.1 uses
thinkingLevel(with settings likeminimal,low,medium, andhigh) instead ofthinkingBudget. The default isminimalto optimize for lowest latency. See Thinking levels and budgets. - Server events : A single
BidiGenerateContentServerContentevent can now contain multiple content parts simultaneously (for example, audio chunks and transcript). Update your code to process all parts in each event to avoid missing content. - Client content :
send_client_contentis supported throughout the entire session lifecycle with explicit roles (userormodel). Settingturn_complete=trueunconditionally interrupts active model generation. See Incremental content updates. - Turn coverage : Defaults to
TURN_INCLUDES_AUDIO_ACTIVITY_AND_ALL_VIDEOinstead ofTURN_INCLUDES_ONLY_ACTIVITY. The model's turn now includes detected audio activity and all video frames. If your application currently sends a constant stream of video frames, you may want to update your application to only send video frames when there is audio activity to avoid incurring additional costs. - Async function calling : Not yet supported. Function calling is synchronous only. The model will not start responding until you've sent the tool response. See Async function calling.
- Proactive audio and affective dialogue : These features are not yet supported in Gemini 3.1 Flash Live. Remove any configuration for these features from your code. See Proactive audio and Affective dialogue.
For a detailed feature comparison, see the Model comparison table in the capabilities guide.
[*] ↩︎