Gemini 3.5 Flash and Omni Take the Main Stage
At Google I/O 2026, Gemini 3.5 Flash and the Gemini Omni model emerged as the core of Google’s AI story. Gemini 3.5 Flash is positioned as a faster, more efficient model built for advanced agentic workflows, coding assistance, and longer multi-step tasks. Its earlier leak as “Gemini 3 Fast” hinted at an emphasis on speed, now confirmed with a formal launch aimed at high-throughput, developer-heavy use cases. Gemini Omni, already rolling into Google Flow and the Gemini app, anchors Google’s multimodal ambitions by unifying text, image, and video generation within a single system. On stage, Omni powered rich, generative graphical answers and sophisticated media transformations, framing it as the next-generation successor in the Gemini lineup. Together, Flash and Omni redefine the spectrum from latency-sensitive agents to rich, multimodal experiences that developers can start building against immediately.
Inside the Gemini Omni Model: Multimodal by Default
The Gemini Omni model is designed as a true multimodal engine, reshaping how developers think about intelligent features. Live demos showed Omni performing complex video editing: altering scenes, changing environments, and inserting new characters while preserving original motion, all from a standard clip. It also generated multiple video variations from a single photo, shifting angles and visual styles without new footage. Another example focused on automatically adding cinematic effects to a simple video of a woman playing guitar, underscoring Omni’s role as a creative co-pilot. Beyond content creation, Omni drives generative graphical answers inside Gemini, giving users visual-rich responses to text prompts. For developers, the implication is clear: apps no longer have to bolt separate models onto different media types. Omni encourages unified workflows where text, images, video, and potentially future modalities are handled through one consistent, API-driven model.
Gemini 3.5 Flash and Spark: Building Agentic Workflows at Scale
Gemini 3.5 Flash is tailored for the emerging class of agent-first applications. Google framed it as a faster AI model optimized for agentic workflows, coding help, and extended multi-step reasoning, filling the gap between lightweight assistants and heavyweight, multimodal engines like Omni. On desktops, the new Gemini app introduces Spark, an agent mode that works with local folders, connectors, and skills. Spark was demonstrated handling planning tasks across Gmail, Drive, Sheets, and Slides in the background, and a public roadmap teases more proactive behaviors, such as automatically setting up services like delivery orders. Combined with Flash’s speed, Spark turns Gemini into an orchestration layer for everyday work: parsing files, coordinating tools, and executing actions. For developers, this opens a path to build agents that operate across cloud documents, local resources, and external APIs, without stitching together a patchwork of disparate services.
Gemini UI Redesign and New AI Developer Tools
Beyond models, Google I/O showcased a broad push to streamline how developers interact with Gemini. A Gemini UI redesign, already rolling out across mobile and web, embraces a new Liquid Glass aesthetic while also aiming to make multimodal interactions more intuitive. Google AI Studio is expected to gain a dedicated mobile companion, allowing developers to experiment with prompts and write code directly from their phones. On desktop, the Gemini app’s Spark agent integrates with local folders and connectors, turning everyday environments into testbeds for AI-driven workflows. Antigravity 2.0, described as a standalone AI development platform, now generates multiple assets simultaneously, targeting teams building more sophisticated AI applications. Together, these AI developer tools shorten the path from prototype to production, encouraging experimentation not just with model prompts but with full end-to-end AI experiences embedded into existing products.
From Glasses to Commerce: Practical Use Cases Developers Can Target
Google’s announcements highlighted concrete scenarios that point to where developers can take Gemini next. AI glasses demos used Gemini to interpret what the wearer saw, control music, and handle coffee orders while the phone stayed in a pocket, with audio-based smart glasses confirmed for launch later. Gemini for macOS showed how mixed files in Finder—like invoices, notes, and images—can be turned into structured outputs such as polished emails via voice commands. In commerce, Universal Cart and the Universal Commerce Protocol illustrated how AI agents can track prices, restocks, compatibility issues, and personalize shopping across Search and the Gemini app. Workspace features like Docs Live and the upcoming Google Pics tool further expand opportunities for document and visual automation. Each demo effectively doubles as a blueprint: developers can adapt similar patterns for their own domains, combining Gemini 3.5 Flash, Omni, Spark, and the revamped UI to build domain-specific agents and multimodal workflows.