Gemini Docs

Live API

• New: Thinking in the Live API covers background reasoning with gemini-3.8-live-extended-thinking, conversational fillers, interaction_status tracking (IN_PROGRESS / IDLE), and guidance on when to use standard Live vs Extended Thinking.
• New model pages: Gemini 3.8 Live (gemini-3.8-live) and Gemini 3.8 Live Extended Thinking (gemini-3.8-live-extended-thinking) are documented as stable Live API audio-to-audio models, including capabilities, limits, and migration/upgrade notes.
• Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking are marked generally available (GA) in the changelog.
• The Live API overview no longer describes the Live API as Preview.
• Live API capabilities now compares Gemini 3.8 Live, Gemini 3.8 Live Extended Thinking, and Gemini 3.1 Flash Live Preview (thinking, client content, async function calling, and related behavior).
• Get started and related guides (SDK, WebSocket, session management, tools, ephemeral tokens) use gemini-3.8-live in place of gemini-3.1-flash-live-preview.
• Best practices notes automatic spoken-language detection (no explicit language code) and that proactive audio is permanently enabled on the Gemini 3.8 Live models.

Models

• Models lists Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking as new stable Live API options; Gemini 3.1 Flash Live is described as a legacy preview model with a recommendation to move to Gemini 3.8 Live.
• Gemini 3.1 Flash Live Preview is documented as legacy, with a pointer to Gemini 3.8 Live and updated client-content behavior notes.

Deprecations

• Deprecations adds gemini-3.8-live and gemini-3.8-live-extended-thinking (no shutdown date announced) and updates several recommended replacements from gemini-3.1-flash-live-preview to gemini-3.8-live.