At Google I/O 2026, Google introduced Gemini 3.5 Flash, the first model in its next-generation Gemini 3.5 family. Designed for frontier-level intelligence and speed, Gemini 3.5 Flash delivers strong performance in coding and agentic tasks. According to Google, it outperforms Gemini 3.1 Pro on challenging coding and agentic benchmarks. Measured by output tokens per second, it is also four times faster than other frontier models.
This update brings advanced AI capabilities closer to real-world deployment and expands the possibilities for AI-powered home devices and applications, accelerating the evolution of the core technologies behind the AI Home experience.
Cedar, SDMC’s home AI agent, has a clear goal: to help homes better understand the people. It combines visual perception, proactive intelligence, household member recognition, and personalized memory. With Gemini 3.5 Flash integrated, Cedar strengthens its reasoning, responsiveness, and context-handling capabilities to support a smarter, faster, and more personalized home experience.
An AI-powered home rarely relies on a single signal. It needs to interpret information across devices, environments, user preferences, and real-time context.
With Gemini 3.5 Flash’s multimodal understanding and agentic capabilities, Cedar can better interpret complex situations and coordinate connected devices more effectively. When a user comes home, for example, Cedar can consider local weather, indoor conditions, and personal preferences, then adjust lighting, air conditioning, and other compatible devices to create a more comfortable environment.
Cedar can also interpret and summarize visual data from connected devices, supporting use cases like alarm summaries, pet status updates, and security event reviews.
Proactive service only matters when it responds in time. Gemini 3.5 Flash’s faster output helps Cedar shorten the path from perception to action, bringing connected home interactions closer to real time.
Whether Cedar is responding to an unfamiliar visitor, retrieving a live camera feed, or coordinating multiple devices, faster model output can make interactions feel smoother and more immediate.
As context capabilities improve, Cedar can enable more consistent use of historical context during longer interactions.
This allows Cedar to maintain a more continuous understanding of household preferences over time. With user consent and appropriate privacy controls, it can retain and recall content preferences, environmental habits, and family routines more reliably. This reduces repeated setup and helps Cedar deliver services that better match individual needs.
Technology will continue to evolve. Cedar’s goal remains clear: to help every home better understand the people who live in it.
Looking ahead, SDMC will continue to work with global customers and ecosystem partners to bring these capabilities to more homes, enabling more users to benefit from AI in everyday life.
©2003-2026 SDMC Technology Co., Ltd
