ToolsSeptember 15, 2026via Vercel Blog
Gemini 3.8 Live models now available on AI Gateway
Why it matters
Google's latest models (including a new extended-thinking variant) are shipping through a key developer platform with unified API, usage tracking, and failover — lowering friction for teams building voice-first and agentic applications.
Key signals
- Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking now available via Vercel AI Gateway
- Models support real-time audio interactions and voice assistants
- Extended Thinking variant adds multi-step reasoning in parallel with speech
- Gemini 3.8 Live supports 97 languages, automatic language switching, and background tool calls
- Access via realtime API with WebSocket client; short-lived token auth
- AI Gateway provides unified API, usage/cost tracking, retries, failover, performance optimization
- Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking now on AI Gateway
- Real-time audio I/O support for voice assistants and conversational experiences
- Extended Thinking variant: multi-step reasoning in parallel with speech, can acknowledge + narrate progress
- Gemini 3.8 Live supports automatic language switching across 97 languages, background tool calls
- Access via AI SDK realtime API; WebSocket-based; short-lived token auth
- AI Gateway layer provides unified API, usage tracking, retries, failover, performance optimization
- Available in model playground for testing
The hook
Gemini 3.8 Live models now available on Vercel's AI Gateway — real-time voice and extended reasoning for production apps.
and from Google are now available on AI Gateway.Gemini 3.8 LiveGemini 3.8 Live Extended Thinking
Both models support real-time spoken interactions for voice assistants, conversational experiences, and applications that respond through audio.
Use either model through the AI SDK's realtime API. Insta…