← Models

Gemini 3.5 Live Translate

Google · Gemini 3.5

Google's realtime speech-to-speech translation model for the Gemini Live API.

Part of Gemini Flash family · Other versions: Gemini 3.5 Flash , Gemini 3.1 Flash-Lite , Gemini 3.1 Flash TTS Preview , Gemini 3.1 Flash Live Preview , Gemini 3 Flash , Gemini 2.5 Flash Live Preview , Gemini 2.5 Flash-Lite , Gemini 2.5 Flash
Type
audio
Context
131K tokens
Max Output
66K tokens
Status
preview
Input
$3.5/1M tok
Output
$21/1M tok
API Access
Yes
License
proprietary
live-api translation speech-to-speech realtime audio preview
Released June 2026 · Updated July 6, 2026

Overview

Freshness note: Live API model behavior, pricing, and availability can change quickly. This profile is a point-in-time snapshot last verified on July 6, 2026.

Gemini 3.5 Live Translate is Google’s low-latency speech-to-speech translation model for the Gemini Live API. The API model ID is gemini-3.5-live-translate-preview, and Google describes it as a real-time translation route for spoken conversations across more than 70 languages.

The mental model is different from a live assistant. This model is an interpreter pipeline: stream audio in one language, receive translated audio and transcript output in another. It is not designed to reason, use tools, or act on the user’s behalf.

Capabilities

Gemini 3.5 Live Translate accepts audio speech input and returns translated speech plus text transcript output. It supports the Gemini Live API and continuous stream processing, so it can translate while the speaker talks instead of waiting for classic turn boundaries.

Google’s Live Translation guide explicitly separates it from Live Agent behavior. Live Agent can use tools, instructions, and multimodal interaction patterns. Live Translation is audio-restricted and translation-only, with configuration focused on target language and related translation settings.

Technical Details

Current public model details:

  • Model ID: gemini-3.5-live-translate-preview
  • Input: audio speech
  • Output: translated audio and text transcript
  • Input token limit: 131,072
  • Output token limit: 65,536
  • Live API: supported
  • Latest update: June 2026
  • Knowledge cutoff: January 2025

Notable unsupported capabilities include audio generation outside the translation output, tools, function calling, Google Search grounding, Google Maps grounding, structured outputs, URL context, thinking, Batch API, Flex inference, and Priority inference.

Pricing & Access

Google lists Gemini 3.5 Live Translate on the Gemini API free and paid tiers. Current paid-tier pricing is:

  • Audio input: 3.50per1Mtokens,orabout3.50 per 1M tokens, or about 0.0053 per minute
  • Audio output: 21.00per1Mtokens,orabout21.00 per 1M tokens, or about 0.0315 per minute
  • Effective combined audio token price: about $0.0368 per minute

Because this is a Live API preview model, production teams should expect tighter operational constraints and more change than on stable text routes.

Best Use Cases

Choose Gemini 3.5 Live Translate for live interpreting, multilingual meeting assistance, language-learning drills, contact-center experiments, guided tours, and conversational support where real-time voice translation matters.

Do not use it as a general voice agent. If the system needs tool use, instructions, multimodal context, or user-goal execution, use a Gemini Live agent model instead and design the translation workflow separately.

Comparisons

  • Gemini 3.1 Flash Live Preview: Better for realtime voice agents and multimodal dialogue; Live Translate is narrower and translation-specific.
  • Gemini 3.5 Flash: Better for text, coding, search-grounded work, and general agent loops; Live Translate is for low-latency audio translation.
  • Speech-to-text plus text translation stacks: More modular and auditable, but usually less natural for bidirectional real-time conversation.