Gemini 3.8 Live and 3.8 Live Extended Thinking

2026-09-15 · Publisher : Google · Domain : Artificial Intelligence

Google launches Gemini 3.8 Live and 3.8 Live Extended Thinking, two near-real-time voice dialogue models, the second reasoning while speaking, for voice agents, Search Live, Gemini Live and Workspace.

What changes

3.8 Live: fluid conversation at controlled cost, near-real-time visual input, 97 languages with language switching mid-conversation.
3.8 Live Extended Thinking: multi-step reasoning during speech, verbal cues and progress narration.
Execution of tools and API calls in the background without interrupting the conversation.
Ranked first on the Artificial Analysis Speech to Speech index for Extended Thinking (82.6).
3.8 Live with Live Avatar variant (animated video avatar) available to enterprises on September 24, 2026.

### Specifications - Availability: Gemini API and AI Studio; Search Live; Gemini Live; Workspace (Docs Live, Gmail Live, Keep Live); private preview on Gemini Enterprise. - Identifiers: gemini-3.8-live and gemini-3.8-live-extended-thinking. - Audio output marked with SynthID. ### Pricing Current pricing page (grouped line for 3.8 Live and 3.8 Live Extended Thinking): input 0.75 USD per million tokens (text) or 0.005 USD per minute (audio); output 4.50 USD (text) or 0.018 USD per minute (audio). ### Benchmarks | Benchmark (scores published by the vendor) | 3.8 Live Extended Thinking | |---|---| | Artificial Analysis Speech to Speech Quality Index | 82.6 | | τ-Voice (agentic tasks) | 68.6% | | τ-Voice-banking (Sierra) | 35.1% | | Big Bench Audio | 97.7% | ### Key takeaways A new generation of voice models for customer service agents, with asynchronous tool calls.

Links

Tags : Audio, Google, Gemini, Live, Voix