Google launches Gemini 3.8 Live and 3.5 Transcribe models
Google added Gemini 3.8 Live, 3.8 Live Extended Thinking, and Gemini 3.5 Transcribe to its API, expanding real‑time voice development tools.
Original source published: September 15, 2026
Google announced that Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking are now available through the Gemini API and Google AI Studio. The models support asynchronous function calling, visual context, alphanumeric precision, multilingual coverage for 97+ languages, and incremental content updates. Extended Thinking adds configurable reasoning for complex, multi‑step tasks. Pricing is listed at $0.005 per minute for audio input and $0.018 per minute for audio output, and the models can be accessed via the Live API and integration partners such as Agora, Fishjam, LiveKit, LangChain, Pipecat, Vercel, and Vision Agents. Gemini 3.5 Transcribe, released the previous month, provides streaming transcription with a 4.0% word‑error rate (2.6% non‑streaming) across 85+ languages, automatic code‑switching, custom vocabulary biasing for up to 1,000 terms, and a smart transcription mode that formats and cleans the output. These additions give developers new building blocks for voice‑first applications, real‑time captioning, and audio analytics.