Gemini 3.8 Live

Gemini 3.8 Live (15 September 2026) is Google's live dialogue model for voice agents — built for near real-time conversation rather than turn-by-turn chat, with visual grounding and the ability to handle being interrupted.

Best for

• Voice agents that must survive a real conversation
• Support and sales lines where the caller talks over the bot
• Multilingual dialogue, including switching language mid-sentence
• Products where latency matters more than the last few benchmark points

How it compares — and the India angle

• The interesting part for India is the language switching: a caller moving between Hindi and English in one sentence is the normal case here, not an edge case, and it is exactly where scripted voice bots fall apart.
• It is the cost-efficient half of the pair — Gemini 3.8 Live Extended Thinking is the one to reach for on multi-step tasks.
• Available to developers in the Gemini API and Google AI Studio, in private preview for Gemini Enterprise, and to everyone through Search Live.
• Google has not published per-token pricing for the Live models on its public pricing page, so none is quoted here — check the API pricing page before committing a budget.