Overview
A preview native-audio model for the Gemini Live API, designed for low-latency, real-time voice and video interaction. Google says the December 2025 update improves handling of complex workflows. It can be used to build conversational voice agents and applications that respond to continuous audio, video or text streams. (aistudio.google.com)
Capabilities
- Accepts audio, video and text input; produces audio and text output.
- Supports audio generation, function calling, Google Search grounding, thinking and the Live API.
- Google describes the audio output as optimized for pacing, naturalness, expressiveness and mood. (aistudio.google.com)
Limits and availability
- Input limit: 131,072 tokens; output limit: 8,192 tokens. Knowledge cutoff: January 2025.
- Context caching, code execution, file search, Google Maps grounding, image generation, structured outputs and URL context are not supported. Batch API, Flex inference and Priority inference are also not supported. (aistudio.google.com)
- This is a preview model. Google's lifecycle page lists a shutdown date of March 31, 2027 and recommends
gemini-3.8-liveas a replacement. (aistudio.google.com)
Pricing
Google's Gemini Developer API pricing, checked November 17, 2026, lists the following paid-tier rates in USD per 1 million tokens:
| Direction | Modality | Price |
|---|---|---|
| Input | Text | $0.50 |
| Input | Audio or video | $3.00 |
| Output | Text | $2.00 |
| Output | Audio | $12.00 |
Output pricing includes thinking tokens. The listed free tier has no input or output charge. These are Gemini Developer API rates for this model, not prices for other platforms or model versions. (aistudio.google.com)
Release and official sources
- Released: December 12, 2025. Google release notes. (aistudio.google.com)
- Model documentation.
- Gemini Developer API pricing, checked November 17, 2026.
- Model lifecycle and shutdown dates.