EmbeddingGemma 2: an open, lightweight multimodal embedding model
deepmind.google·6d ago
AI Engineering
Hacker News·27d ago·5 min read
Gemini 3.8 Live and 3.8 Live Extended Thinking
TL;DR
Gemini 3.8 Live and 3.8 Live Extended Thinking have been launched, enhancing voice agents with real-time reasoning and intuitive dialogue capabilities. These models enable more efficient and intelligent interactions across various platforms.
✦ Why It Matters
Explore integrating Gemini 3.8 models into your voice applications for enhanced user experience.
Key Takeaways
- Integrate Gemini 3.8 Live for improved voice agent performance.
- Utilize 3.8 Live Extended Thinking for complex task automation.
- Leverage real-time visual inputs for enriched conversations.
- Implement SynthID watermarking to ensure content traceability.
Full Summary
Gemini 3.8 Live and 3.8 Live Extended Thinking are new AI models designed to improve voice interactions by enabling near real-time reasoning. Gemini 3.8 Live focuses on conversational intelligence and visual grounding, while 3.8 Live Extended Thinking is tailored for complex tasks requiring multi-step reasoning.
The models excel in performance metrics, with 3.8 Live Extended Thinking achieving top scores in speech quality and task completion benchmarks. They support 97 languages and can execute background tasks while maintaining conversation flow.
Developers can integrate these models via the Gemini API, enhancing user experiences across platforms like Google Workspace and Search. The introduction of SynthID watermarking ensures that AI-generated audio remains traceable, addressing misinformation concerns.