Google Unveils Gemini 3.5, Spark Agent and Omni World Model at I/O 2026
Google used its annual developer showcase on Tuesday to unveil a wave of AI products, CNBC reported, including a faster model, a personal AI agent, and a world-simulation tool designed to keep the search giant competitive with OpenAI and Anthropic.
Gemini 3.5 Flash Takes Center Stage
The headline product is Gemini 3.5 Flash, a lighter but capable addition to Google’s model lineup. CEO Sundar Pichai told reporters ahead of the Google I/O event that the model is “remarkably fast.” Google says it delivers frontier-level performance at roughly half the cost of comparable models, and in some cases closer to one-third. Gemini 3.5 Flash will now serve as the default model inside the Gemini app and Google’s AI search mode globally. Google also highlighted improved safety guardrails on the model. The company said it is less prone to generating harmful content and less likely to wrongly refuse legitimate queries. A heavier version, Gemini 3.5 Pro, is being tested internally and is expected to reach wider audiences next month.
A Personal Agent Called Spark
On the agentic side, Google introduced Gemini Spark, a general-purpose AI agent built into the Gemini app. Spark is designed to reason across a user’s connected applications and take actions on their behalf. Google described the goal as helping users manage their digital lives with minimal hands-on effort. Spark launches in beta next week, available first to trusted testers and Google AI Ultra subscribers. The move reflects Wall Street pressure on Alphabet to demonstrate that its heavy capital spending translates into meaningful product integrations across its ecosystem.
Background: A Race With High Stakes
Google’s announcements come at a moment of intense scrutiny across the AI sector. Anthropic recently released its Mythos model, which researchers said uncovered thousands of previously unknown software vulnerabilities. Both Anthropic and OpenAI are preparing for potential initial public offerings as early as this year, drawing investor attention and raising the bar for incumbents like Google.
Omni Brings World Modeling to Consumers
Rounding out the announcements is Omni, a world model developed through years of DeepMind research. Unlike conventional generative tools, Omni is built to simulate physical environments and predict outcomes based on user actions. Google said Omni will integrate with Gemini Flash, the Gemini app, Google Flow, and YouTube Shorts. The model supports image and audio inputs and lets users modify video footage by altering actions, adding characters, or inserting objects into scenes. The breadth of Tuesday’s launches signals Google’s intent to compete not just on model benchmarks but on the depth and reach of its AI-powered product suite.
Read Next: OpenAI Eyes IPO as Valuation Surges Past $300 Billion
