Gemini 3.8 Live Extended Thinking
Google's live voice model for conversations that require multi-step reasoning and background tool use.
Overview
Google introduced Gemini 3.8 Live Extended Thinking on 15 September 2026. The announcement describes a voice model that can reason while speaking and continue a conversation during background tasks. Google reports a launch score of 82.6 on Artificial Analysis' Speech to Speech Quality Index; this is a reported benchmark result, not an Aymorphic evaluation.
This is a sourced directory entry. Features and capabilities are reported in the linked sources; this entry does not document independent Aymorphic testing.
Aymorphic analysis
Evaluate this model on interrupted conversations and unfinished tool calls, not only answer quality. A booking assistant needs to preserve task state when a user changes a request halfway through.
Limitations
Desk review of Google's announcement. No independent voice, latency or benchmark testing was performed.
Key features
- Concurrent speech and reasoning
- Background tool calls
- Multilingual live dialogue
Use cases
- Voice assistants with multi-step tasks
- Conversational troubleshooting
Pricing
Check current Gemini API pricing and access terms; availability differs by product and subscription.
Pricing reflects the source record and may change. Confirm current terms with the provider.
Dates and sources
- Released
- 15 Sept 2026
- Record updated
- 16 Oct 2026
- Record last checked
- 16 Oct 2026