Google · voice model

Gemini 3.8 Live

Gemini 3.8 Live is a Google voice model released Sep 15, 2026; Google's newest voice model in our dataset is Gemini 3.8 Flash TTS. Proprietary, API access only. Google's real-time voice model for the Gemini Live API, generally available since Sep 15, 2026, with a separate Extended Thinking variant. Verified against vendor sources on Oct 26, 2026.

Reported Last verified Oct 26, 2026 · All models · JSON · Markdown

ReleasedSep 15, 2026
LicenceProprietaryProprietary (API)

InTextAudioOutTextAudio

Key facts

Gemini 3.8 Live: key facts with sources
Fact	Value	Source
Vendor	Google	n/a
Released	Sep 15, 2026	source
Status	generally available	n/a
API id	gemini-3.8-live	source
Input	text, audio	source
Output	text, audio	source
Licence	Proprietary (API)	source
Pricing	Text input $0.75; Audio input $3; Text output $4.5; Audio output $12 (per 1M tokens)	source

Where Gemini 3.8 Live sits in our data

Released Sep 15, 2026 (week 38 of 2026), in September 2026. Other models in our data released that month: GPT-6.1 Sol, Claude Sonnet 5.5, Claude Opus 5.5, Gemini 3.8 Flash TTS, Gemini 3.8 Flash-Lite TTS, GPT-6 Luna.
The previous Google model we track is Gemini 3.8 Flash (Sep 2, 2026), 13 days earlier.
Google tools we cover: Gemini Notebook, Google ADK, Gemini CLI.
All releases by date: AI model releases by month · new AI models this week.

How Gemini 3.8 Live differs from Gemini 3.8 Live Extended Thinking

Read from the recorded facts of the two models, not from a head-to-head test. We have not run either.

We have a recorded price for Gemini 3.8 Live but none for Gemini 3.8 Live Extended Thinking.
Input is text, audio, against text, image, audio for Gemini 3.8 Live Extended Thinking.
Both were released on Sep 15, 2026.
Facts for the other model: Gemini 3.8 Live Extended Thinking.

Notes

The Extended Thinking variant is gemini-3.8-live-extended-thinking (same release date).

What we tested

Nothing yet. We have not run Gemini 3.8 Live ourselves, so every fact on this page is Reported from the sources below. When we do test it, the scope will be stated here and in our test logs. See methodology.