Intel / post

Gemini 3.8 Live Tops Speech-to-Speech Benchmarks

Source: ArtificialAnlys
Date: Published Sep 15, 2026

Context

Google DeepMind's Gemini 3.8 Live is a speech-to-speech model that executes tools and API calls in the background while a conversation continues, and it succeeds Gemini 3.1 Flash Live, according to Artificial Analysis's benchmark thread. It comes as a standard model and as Extended Thinking, which supports configurable reasoning effort.

Artificial Analysis tested Extended Thinking at High reasoning effort. It debuts first on the Speech to Speech Index at 82.6, ahead of GPT-Live-1 (Astra, medium) at 81.5 and Grok Voice Think Fast 2.0 High at 81.3, while the standard model is fifth at 76.0. It also leads the Tau Voice agentic at 68.6%, up from 37.7% for Gemini 3.1 Flash Live High. The lead has costs. On the Speech Arena, Extended Thinking trails on preference (Elo 990 versus 1083 for the standard model), its average time to first audio is 1.35 seconds versus 1.18, and it costs $3.50 per hour of input audio versus $0.84.

Terms in this piece · Glossary

benchmark — A standard public test set for comparing AI models — the shared scoreboards behind every "model X beats model Y" claim.
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.

Read the source x.com