Google I/O 2026: Gemini Omni for AI Videos, Gemini 3.5 Series Models Unveiled

Google I/O 2026 was hosted on Wednesday, giving everyone a first look at the new features and products the company will be rolling out to users and enterprises. Google CEO Sundar Pichai kicked off the keynote session and announced new artificial intelligence (AI) tools, such as Docs Live. However. It wasn't until DeepMind CEO Demis Hassabis took the stage that the company revealed its most exciting innovation, Gemini Omni. Separately, the company also introduced the Gemini 3.5 series models for users.

Gemini 3.5 Series, Gemini Omni Unveiled

During the live event, the company unveiled the Gemini 3.5 series. Successor to the Gemini 3.1 series, the latest models bring significant upgrades in agentic capabilities and coding performance. Currently, Google is rolling out the Gemini 3.5 Flash model globally to everyone in the Gemini app and AI Mode in Search. It is also available to developers via the Antigravity platform and in the Gemini application programming interface (API) via the AI Studio and Android Studio.

Google says Gemini 3.5 Flash delivers "frontier-level performance at 4x the speed of comparable frontier models, at less than half the cost." Based on internal evaluations, the company claims that 3.5 Flash outperforms Claude Sonnet 4.6, Claude Opus 4.7, and GPT-5.5 in MCP Atlas and Toolathlon benchmarks for agentic performance, Finance agent v2 for financial analysis, MMMU Pro for multimodal understanding, and MRCR v2 (1 million pointwise) for long context information processing.

The Gemini 3.5 Flash is also said to outperform Claude Opus 4.7 and GPT-5.5 in terms of speed by using 289 tokens per second for output generation. "Gemini 3.5 Flash is our strongest agentic and coding model yet, outperforming Gemini 3.1 Pro on challenging coding and agentic benchmarks," the company said.

Beyond the numbers, the model's biggest USP appears to be agentic performance. The company highlighted that the model was able to access the Antigravity platform to generate a fully functional operating system in just 12 hours by using 93 agents in parallel and less than $1,000 (roughly Rs. 98,000) in API costs.

Coming to Gemini Omni, it is Google's first video generation model that can combine multimodal input in a single prompt. This means a single prompt can feature videos, audio, text, and images, and create anything. While we do not know what "anything" includes, currently, the Omni Flash model is being used to generate videos. It also lets users edit videos via conversations. The model is capable of changing particular elements, including characters, objects, backgrounds, and more.

Gemini Omni Flash is currently rolling out to Google AI Plus, Pro, and Ultra subscribers globally via the Gemini app and Google Flow. It is also rolling out to users on YouTube Shorts and the YouTube Create App starting this week for free.