← Back to Voice models
Google DeepMindNotable
Gemini 3.8 Live Extended Thinking
Released Sep 15, 2026
The higher-reasoning version of Gemini 3.8 Live, for complex multi-step voice tasks. It runs background reasoning at a configurable level while the conversation continues, and powers Gemini Live in the Gemini app.
- Price per minute
- —
- Pricing
- Same table as Gemini 3.8 Live: input text $0.75 / audio $3.00 ($0.005/min) / image-video $1.00 ($0.002/min); output text $4.50 / audio $12.00 ($0.018/min) per 1M tokens, including thinking tokens.
- Inputs
- audio, text, image, video
- Tool calling
- Yes
- Languages
- 97
- Voices
- 30
- Big Bench Audio
- 97.7%
- Access
- Generally available
- Licence
- Proprietary
Strengths & weaknesses
Strengths
- Big Bench Audio 97.7% (AA)
- Configurable background thinking: minimal, low, medium, high
- Asynchronous function calling while audio keeps streaming
- Same per-token price as Gemini 3.8 Live
Weaknesses
- 1.35s time to first audio on AA, among the slowest of the top models
- τ-Voice 68.6%, behind GPT-Live-1 (74.5%) and Grok Voice 3.0 (71.6%) per xAI's comparison table
Notable: frontier-lab clause (also top 10 on AA: #4 on S2S Index at 82.6, #7 on Big Bench Audio). Google's launch post claimed #1 on AA's S2S index at 82.6; AA now ranks it 4th behind Phonic 1, Grok Voice 3.0 and GPT-Live-1. Google also reports τ-Voice-banking 35.1%. The AA listing is 'Gemini 3.8 Live Extended Thinking (High)'. Language and voice counts as for Gemini 3.8 Live.
Links
API model strings
- google
gemini-3.8-live-extended-thinking