← All models

Gemini 3.8 Live Extended Thinking

Published Publisher Googlegemini-3.8-live-extended-thinking

Overview

Google announced Gemini 3.8 Live Extended Thinking on September 15, 2026; the announcement was updated on September 17, 2026. It is a real-time, audio-to-audio Live API model designed for complex, multistep voice interactions that require deeper reasoning. (blog.google)

Capabilities and uses

  • Performs configurable background reasoning at low, medium, or high levels while streaming spoken responses. It can call asynchronous tools and narrate progress without stopping the conversation. (aistudio.google.com)
  • Supports visual context in live conversations. The Google audio model card describes text, audio, image, and video inputs and text and audio outputs; the model-specific API page lists text, image, and audio inputs and text and audio outputs. (storage.googleapis.com)
  • Suitable for voice agents handling multistep diagnostics, travel or booking workflows, coordinated information retrieval, and code or STEM tutoring. Google makes the model available through the Gemini API, Google AI Studio, Gemini app, Gemini Enterprise Agent Platform, and selected Workspace experiences. (ai.google.dev)

Limits and integration

  • The model-specific documentation lists an input limit of 131,072 tokens and an output limit of 65,536 tokens. (aistudio.google.com)
  • Function calling is asynchronous only. A turnComplete message does not necessarily mean background work is finished; clients should check interaction_status. Context caching, Batch API, priority inference, code execution, structured outputs, and image generation are listed as unsupported. (aistudio.google.com)
  • The model may hallucinate and occasionally experience slowness or timeouts. Google’s audio model card gives a knowledge cutoff of January 2025. (storage.googleapis.com)

Gemini Developer API pricing

Currency: USD. The official shared pricing table for Gemini 3.8 Live, Gemini 3.8 Live Extended Thinking, and Gemini 3.1 Flash Live Preview lists the following paid-tier rates, checked November 16, 2026. Token prices are per 1 million tokens: (aistudio.google.com)

Direction Modality Paid-tier price
Input Text $0.75 per 1 million tokens
Input Audio $3.00 per 1 million tokens, or $0.005 per minute
Input Image/video $1.00 per 1 million tokens, or $0.002 per minute
Output Text $4.50 per 1 million tokens
Output Audio $12.00 per 1 million tokens, or $0.018 per minute

Output pricing includes thinking tokens. The same table lists free-tier input and output as free of charge. Google Search grounding has a separate charge: the paid tier includes 5,000 free search requests per month shared across Gemini 3 and newer models, then costs $14 per 1,000 requests, subject to the stated daily cap. These are Gemini Developer API rates, not a claim about pricing on other platforms. (aistudio.google.com)

Official sources