Google unveils Gemini Omni and Gemini 3.5 Flash demos

Google used its I/O 2026 keynote to show how Gemini Omni and the Gemini 3.5 family are meant to work in practice. The company said Omni can create videos from mixed inputs such as images, audio, video and text, while Gemini 3.5 Flash is designed for agentic tasks, coding and long-horizon workflows.

According to Google, Gemini Omni starts with video and lets users edit that video through conversation. The model keeps characters consistent across edits, preserves scene continuity and remembers what happened earlier in the prompt thread. In the demos, Google showed users changing a sculpture into bubbles, altering the lighting in a room, and moving a violinist into a different setting while keeping the original scene intact.

Google also showed a more elaborate Omni example: a prompt that asked for a black-and-white checkerboard room inside a glass sphere, with a recursive image of the same hand holding the sphere and the camera slowly moving closer. The point of the demo was to show that Omni can reimagine action and scene structure instead of only making small edits.

The second major model, Gemini 3.5 Flash, is Google’s latest Flash-series model. Google said it combines frontier intelligence with action, and is built for agentic systems that need to complete multi-step tasks reliably at speed. The company said it is the first model in the 3.5 line and is aimed at long-horizon work that delivers practical value.

In one demo, 3.5 Flash ran multi-step workflows to rename and categorize unstructured assets based on changing criteria. Google said this was powered by Antigravity, its updated harness for coordinating collaborative subagents. In that setup, 3.5 Flash handled multiple steps under supervision while maintaining performance on coding and task execution.

Google also used AI Studio to show 3.5 Flash generating different user interface approaches for a checkout flow in about 60 seconds. The company said the model can build richer web UIs and graphics, drawing on the multimodal foundation of Gemini 3.

Google is already pushing 3.5 Flash into products. The company said it is now the default model for the Gemini app and AI Mode in Search globally. Google said Search will use the model to power new information agents that run in the background, reason across sources, and send updates with links so users can act on them.

Those information agents will first launch for Google AI Pro and Ultra subscribers this summer. Google also said generative UI in Search, including visual tools and simulations tailored to specific questions, will be available for everyone this summer at no charge.

For ongoing tasks, Google said Search will also generate dashboards, trackers and mini apps that users can return to later. The company said custom experiences will be possible through Antigravity in Search in the coming months, first for Google AI Pro and Ultra subscribers in the U.S.

Google introduced Gemini Spark as a personal AI agent built on Gemini 3.5 and the Antigravity harness. The company said Spark runs 24/7, is deeply integrated with Gmail, Docs and Slides, and can take action on a user’s behalf while under direction. Google said Spark is available now to Google AI Ultra subscribers in the U.S.

On availability, Google said Gemini Omni Flash is rolling out to Google AI Plus, Pro and Ultra subscribers globally through the Gemini app and Google Flow, and also at no cost to users on YouTube Shorts and the YouTube Create app. The company said it will roll out to developers and enterprise customers through APIs in the coming weeks. Gemini 3.5 Flash is available through Google Antigravity, the Gemini API in Google AI Studio and Android Studio, Gemini Enterprise Agent Platform and Gemini Enterprise, and is also rolling out globally in the Gemini app and AI Mode in Search.