- Sources: primary, announcement, discussion
- Summary: Google Cloud states that Gemini 3.8 Live with Live Avatar is generally available in Gemini Enterprise, with US and EU endpoints and provisioned throughput. The model carries native speech-to-speech, tool calling during a conversation, 97 languages with automatic detection, and live camera and screen-share understanding, and SynthID watermarks are applied to all generated audio and video. Custom avatars sit behind an enterprise allowlist, and Gemini 3.8 Live Extended Thinking remains in private preview.
- Why it matters: Regional endpoints, provisioned throughput, and watermarking on every generated stream are the terms that decide whether a real-time voice agent can be put into production.
- Follow-up: Track whether Extended Thinking leaves private preview and whether custom avatars open beyond the enterprise allowlist.
send feedback on this story