Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking are Google’s newest real‑time dialogue models, delivering major upgrades in intelligence and parallel reasoning for voice interactions. They handle interruptions, language switches, and background tool execution without breaking the conversation flow.
Gemini 3.8 Live focuses on scale and cost efficiency, offering fluid chats with near‑real‑time visual grounding. It auto‑detects and switches among 97 supported languages and runs tool/API calls in the background while keeping the dialogue alive.
Gemini 3.8 Live Extended Thinking is built for high‑complexity tasks, providing stronger multi‑step reasoning. When it receives cues like “let me check that…”, it acknowledges verbally and continues to narrate progress as background tasks complete, giving the impression of continuous thought.
In benchmarks, Extended Thinking tops Artificial Analysis’s Speech‑to‑Speech Quality Index with an 82.6 score, achieves 68.6% task completion on τ‑Voice, and 35.1% on Sierra’s τ‑Voice‑banking benchmark. It also scores 97.7% on Big Bench Audio while remaining competitively priced. Gemini 3.8 Live placed second in the Speech Agent Arena, balancing high user preference with cost‑effectiveness.
The models are available via the Gemini API, Google AI Studio, Google Workspace, Search, and the Gemini app. Developers can integrate the Live API through platforms such as Agora, Fishjam, LangChain, LiveKit, Pipecat, and Vercel, enabling rapid creation of high‑performance voice‑driven interfaces. Enterprise partners like Salesforce, Genspark, and Lumeris are already testing the models for low latency, fluid interaction, and robust tool‑calling.
All generated audio includes an imperceptible SynthID watermark, ensuring AI‑generated content remains detectable and helping to curb misinformation.
Access points
- Developers: Gemini API & Google AI Studio
- Enterprises: Private preview on Gemini Enterprise, later rolling out to Customer Experience and Workspace
- General users: Search Live, Gemini app, Workspace (Docs, Gmail, Keep)
Through these channels, users can edit documents, schedule meetings, generate code, and complete other complex workflows simply by speaking, truly realizing “talk‑to‑complete”.
Review