Google Releases Gemini 3.8 Family Including Live and Extended Thinking Variants
Something you can actually use or run today.
On September 15, 2026, Google announced the release of Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, models optimized for real-time interaction and deep technical reasoning.
Accelerates low-latency voice interaction and tree-search technical engineering queries within the Google Cloud ecosystem.
Google is segmenting its models by specialized capabilities rather than just size. The Extended Thinking variant directly targets OpenAI's reasoning strengths, preserving Google's relevance among engineering teams.
Direct benchmark comparisons between Gemini 3.8 Extended Thinking and OpenAI's current reasoning models.
Also covers
- Google Releases Gemini 3.8 Family Including Live and Extended Thinking VariantsOn September 15, 2026, Google officially launched the Gemini 3.8 model family, which includes Gemini 3.8 Live for low-latency conversational AI and Gemini 3.8 Live Extended Thinking for complex algorithmic reasoning.
- Accelerates real-time audio interaction loops by reducing duplex speech processing lag.
- Improves success rates on multi-step engineering queries via dedicated tree-search reasoning tracks.
Google is prioritizing real-time voice latency improvements along with analytical thinking architectures.