Gemini 3.8 Live and 3.8 Live Extended Thinking
10 hours ago
- Introduction of Gemini 3.8 Live and 3.8 Live Extended Thinking models for near real-time reasoning and voice agents.
- Gemini 3.8 Live is built for scale and cost efficiency with fluid dialogue and visual grounding.
- Gemini 3.8 Live Extended Thinking is designed for high-complexity tasks with multi-step reasoning and increased intelligence.
- Extended Thinking achieved #1 on Artificial Analysis' Speech to Speech Quality Index and leads in agentic task completion benchmarks (τ-Voice, τ-Voice-banking).
- Live model ranks #2 in Speech Agent Arena and is highly cost-effective, processing visual inputs in near real-time with support for 97 languages and background tool execution.
- Both models offer seamless conversational flow, with Extended Thinking using natural verbal cues and live progress narration for complex tasks.
- Available via Gemini Live API for developers, with integrations from platforms like Agora, LangChain, LiveKit, and partnerships with Salesforce, Genspark, Lumeris.
- All audio outputs are watermarked with SynthID for transparency and misinformation prevention.
- Rollout begins today: 3.8 Live and Extended Thinking are available in Gemini API, Google AI Studio, and for enterprises in private preview.