Skip to content

Google · Voice

Gemini 3.8 Live

Google 🇺🇸VoiceGemini familyOne of its sources says the weights aren't publicly available.
From Empiriolabs's listing:
Low-latency voice conversation over one socket, with 30 voices, live video input, tool calling, and replies you can interrupt as you speak.
Max context
131K
Listings
4
Providers
4

Evidence and sources

Released 15 Sep 2026Maker HQ: US (according to Google's own website)
Sources: EmpirioLabs data, Models.dev data, llminfo researchPrimary data from EmpirioLabs from Models.dev: weights availability from llminfo research: release date

Input & output

as stated by the model's providers
Text
Images
Audio
Video
Text
Audio

Capabilities

Supported operations and provider-stated features

VisionFunction callingWeb searchAudio input

Listings

 Googlegemini-3.8-liveSource: LiteLLM data131K$0.75$4.50$3.00$12.00$1.00$0.000033
 Vercelgoogle/gemini-3.8-liveSource: Models.dev data131K$0.75$4.50————
 Empiriolabsgemini-3-8-liveSource: EmpirioLabs data131K$2.60$11.70————

1 listing without a price is hidden.

Benchmarks

Named effort levels run from lowest to highest. Select a result to see its source details.

Speech

Artificial Analysis3 related benchmarks · 4 results
τ-Voice · Agentic Performance
Reported by Google DeepMind · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
Gemini API · pass@1 · thinking level not printed on chart
Input Audio Cost on Big Bench Audio Subset
Reported by Google DeepMind · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
Gemini API · Big Bench Audio subset · thinking level not printed on chart
Speech to Speech Quality Index
Reported by Artificial Analysis · via BenchBoard docs
Artificial Analysis live leaderboard · October 24 snapshot · index methodology as displayed by maintainer
Reported by Google DeepMind · via BenchBoard docs · Reported by the model's developer. The source relaying it did not verify it.
Gemini API · Artificial Analysis composite · thinking level not printed on chart

Data credits: BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links.

Around the web

Also known as: gemini-3-8-live, gemini/gemini-3.8-live, google/gemini-3.8-live, vertex_ai/gemini-3.8-live

Like this one. Tap or hover any dotted value, label or icon to see where it comes from and how it is measured. open details when tapped or hovered. How this data works