← All models

Gemini 3.8 Live

Published Publisher Googlegemini-3.8-live

Overview

Gemini 3.8 Live is a low-latency, audio-to-audio model for real-time dialogue and voice agents. Google positions it for scalable, cost-efficient conversational applications with visual grounding. It is distinct from Gemini 3.8 Live Extended Thinking, which is designed for more complex, multi-step reasoning. (blog.google)

Capabilities and uses

  • Supports interleaved reasoning, streaming audio, synchronous and asynchronous function calling, and Google Search grounding. Background tool calls can continue while the conversation proceeds. (aistudio.google.com)
  • Can use near-real-time visual context and switch between 97 supported languages during a conversation, according to the launch announcement. Example uses include voice support, employee onboarding, troubleshooting, and conversational agents that call external tools. (blog.google)
  • The Gemini API model documentation lists text, image, and audio inputs and text and audio outputs. Its Live API guidance specifies audio as the response modality, with optional output transcription; it also discusses sending video frames. Platform capabilities may differ. (aistudio.google.com)

Limits and availability

  • Gemini API limits are 131,072 input tokens and 65,536 output tokens. Context caching, code execution, file search, image generation, structured output, URL context, Google Maps grounding, Batch, Flex, and Priority inference are listed as unsupported. (aistudio.google.com)
  • The base model does not support configurable thinking_level; Google directs applications requiring configurable background reasoning to the separate Extended Thinking model. Proactive audio is always enabled in the documented Live API configuration. (aistudio.google.com)
  • Google announced and released the models through the Gemini API and Google AI Studio on September 15, 2026. Google Cloud separately lists the model's general-availability release on its platform as September 24, 2026. (blog.google)

Pricing

Gemini Developer API, paid tier, USD, checked November 16, 2026. Prices are per 1 million tokens unless otherwise stated: (aistudio.google.com)

Direction Modality Price
Input Text $0.75
Input Audio $3.00; Google also lists $0.005/minute
Input Image/video $1.00; Google also lists $0.002/minute
Output Text, including thinking tokens $4.50
Output Audio $12.00; Google also lists $0.018/minute

The pricing page lists model input and output as free on the free tier, subject to applicable availability and limits. Paid-tier Google Search grounding includes 5,000 free search requests per month shared across Gemini 3 and newer models, then $14 per 1,000 search requests, subject to the listed daily cap. A model request can generate multiple billable search queries. (aistudio.google.com)

For comparison, Google Cloud Gemini Enterprise Agent Platform, non-global standard pricing, lists the same text, image/video input, audio input, text output, and audio output token rates above. It separately lists avatar video output at $1.00 per 1 million tokens; this is a platform-specific capability and charge. Checked November 16, 2026. (cloud.google.com)

Official sources