Google used its annual I/O developer conference to roll out a wave of AI products and upgrades, including the next generation of its flagship Gemini model, a new family of multimodal Omni models, and a revamped AI-powered Search experience. The event, held at the Shoreline Amphitheatre, was dominated by the company’s push to embed artificial intelligence across its entire ecosystem, from its core search engine to hardware devices.
Gemini 3.5 Takes Center Stage
The star of the show was Gemini 3.5, Google’s latest large language model. It succeeds last year’s Gemini 2.0 and promises significant improvements in reasoning speed, context retention, and multimodal understanding. Google demonstrated Gemini 3.5 processing long-form video, audio, and text simultaneously, answering complex questions in near real-time. The company says the model will roll out to developers and consumers in phases, with a free tier and paid Pro version.
New Omni Models for Multimodal Tasks
Alongside Gemini 3.5, Google introduced a new family of Omni models. These models are designed to handle inputs across text, image, audio, and video with what the company described as “seamless fusion.” Unlike earlier multimodal systems that often juggled separate components, Omni models process all modalities in a single trained architecture. Google said the Omni models will be available through Vertex AI for enterprise customers first, with consumer applications expected later this year.
AI Search Gets a Smarter Overhaul
Google also revamped its AI-powered Search. The update integrates Gemini 3.5 directly into the search ranking and snippet generation systems. Users will see more detailed, multi-step answers for complex queries—like trip planning or product comparisons—without clicking on multiple links. The revamped search includes live fact-checking prompts that cite sources explicitly. Google said the new search experience will begin rolling out in the US this week and expand globally over the next quarter.
New Hardware to Run AI Locally
On the hardware side, Google unveiled a new device lineup: the Pixel 11 smartphone and the Nest Home Max 2 smart speaker, both built with custom Tensor G6 chips that handle AI processing on the device rather than in the cloud. The Pixel 11 includes a dedicated neural engine that powers on-device Gemini 3.5 Mini, allowing tasks like real-time language translation and photo editing without an internet connection. The Nest Home Max 2 features an upgraded assistant that can recognize multiple voices and maintain context across conversations.
AI Everywhere
Throughout the keynote, Google emphasized that AI is now embedded in every major product it sells—from Workspace apps like Gmail and Docs to YouTube, Maps, and Android. The company showed Gemini 3.5 generating full meeting summaries in Google Meet and editing documents from voice prompts. Developers gained access to new APIs that let third-party apps tap into the same underlying models. Google’s message was clear: AI is no longer a separate feature but the foundation of the entire ecosystem.
What remains unclear is how fast these features will reach users outside the US and what pricing will look like for the advanced models. Google has not yet announced international rollout dates for Gemini 3.5’s premium tier or the full Omni models.
