Google launches Gemini 3.8 Live with extended thinking
Google has launched Gemini 3.8 Live with real-time speech, vision, and extended thinking across 97 languages. The model can run tools and process complex tasks asynchronously in the background while maintaining an unbroken voice conversation with the user.
Natural vocal inflection is now table stakes—the true battleground for voice AI is whether an agent can reason and execute background tasks without breaking conversational flow.
• Decoupling tool execution from voice generation eliminates the unnatural pauses and latency delays that break immersion in current real-time voice assistants.
• Extended Thinking in a live conversational modality represents a significant leap forward, allowing deep reasoning without sacrificing real-time responsiveness.
• Native multi-modal comprehension spanning 97 languages with vision and audio firmly targets global enterprise and consumer adoption ahead of competing voice modes.
• The ultimate test will be task completion reliability: agents must execute real-world workflows without losing state, hallucinating parameters, or dropping conversational context.
DISCOVERED
2h ago
2026-09-17
PUBLISHED
2h ago
2026-09-17
RELEVANCE
AUTHOR
SreeramG
