Google Gemini, Gemma and embodied-AI expansion
Key Questions
What new features are coming to Gemini models?
Gemini 3.5 Flash is now the default model with Omni video live capabilities. Gemini 3.5 Live Translate and Agents CLI were also released.
What updates were made to Gemma 4?
Gemma 4 is an open model that is three times faster with an encoder-free 12B variant. QAT support enables efficient local inference including in-browser runs.
How is Google integrating Gemini with robotics?
Gemini now powers humanoid robots like Atlas and Apollo using split-brain architecture and fleet learning. This allows instant skill propagation across robot fleets.
Google continues broad Gemini deployment across Search, developer tooling, translation, agents, and robotics. Gemini-powered humanoid robots with fleet learning and highly efficient Gemma 4 variants extend the story beyond chat into production software and on-device or embodied systems.